Future-oriented studies

Future-oriented studies

Other titles in this séries New technologies and development Biotechnologies in perspective Culture des îles et développement/Islands' culture and development Biotechnologies in developing countries: présent and future. Vol. l: Régional and national survey

Michel Godet

From anticipation to action A handbook of stratégie prospective Préface by Joseph F. Coates

UNESCO

Publishing

Originally published as De l'anticipation à l'action: manuel de prospective et de stratégie

© Dunod, Paris, 1991 Translated from the Frcnch by Clare Degenhardt Rcvised by Valérie Shepherd The author is rcsponsibie for the choice and the présentation of the facts contained in this book and for the opinions expressed therein, which are not necessarily those of UNESCO and do not commit the Organization. The désignations employed and the présentation of the material throughout this publication do not imply the expression of any opinion whatsoever on the part of UNESCO concerning the légal status of any country, territory, city or area or of iîs authorities, or concerning the délimitation of its frontiers or boundaries. Published in 1994 by the United Nations Educational, Scientific and Cultural Organization 7 place de Fontenoy, 75352 Paris 07 SP, France Composed by Éric Frogé Printed by Presses Universitaires de France^ Vendôme ISBN 92-3-102832-4 ©UNESCO 1994 Printed in France

Préface

In an âge of anxiety and a period of transition., ail institutions crave insight into the future. The globalizing of the world's économies is creating new complexities for business planning and aggravating uncertainties about the business environment. The imminence of the miliennium is, for many, a social milestone, a point at which the future of nations will be seen to be bright and optimistic or in décline. In every sector, stocktaking, anticipation and plans are the order of the day as we approach that transition. This attention to the future is surely favourable to the human enterprise. But the anticipation of the future is stérile unless it is accompanied by systematic, well-grounded, comprehensive actionable plans. Since the séminal work of the Marquis de Condorcet at the tirne of the French Révolution, the Western worid has learned that it can identify long-term trends and that it can anticipate many of the conséquences of new developments. Further, it is clear that our actions influence the future. But more important is the récognition that we can consciously act to influence the future in directions in which we wish it to evolve. Surely no one but a fool would ciaim that we can see the future with crystalline clarity or that we can take actions which will détermine the future. AH we need is to agrée on this modest claim: that we can see the future to a degree that is useful for planning and that we can take actions which will make désirable outcomes more likely and undesirable outcomes less likely. Central to the current look to the future is the concept of alternatives - that in view of ail the complexity ahead of us, it makes little sensé to say 'this will happen' or 'that won't happen1. Taking complexity into

Préface

account, we need to recognize a range of alternative futures as the basis for planning and systematic action. Professor Godet's vvork is an outstanding example of the intégration of anticipation and planning for action. But his strengths as an analyst and as a planner go well beyond that. Too much of the study of the future - even today in this globalizing world - is parochial and provincial. Few British or American futurists, planners and forecasters pay much or any attention to work done outside their national boundaries and even less attention to work done outside their native tongue. One of the accomplishments of Professor Godet is to vvork across cultures and across languages to give his work a value which is enhanced by his catholicity of perspective and broad embrace of materials. By being cross-national in its orientation, Professor Godet's work is an unusuat, if not unique, contribution to the management and futures literature. It draws on a wide range of expérience and illustrations and offers a broader démonstration of techniques practised by corporations and other organizations than is generally found in other sources. Godet's work is an attractive intégration of theory, the historical évolution of management and planning techniques and the practical tools of the trade, ail amply iliustrated. Again, unlike other vvriters in the field, Professor Godet offers a felicitous combination of technique and substance. Beside UNESCO's traditional readership of national and régional planners, the primary audience for Professor Godet's book is business and other organizational planners. The secondary, and nearly as important, audience is professors teaching courses in business schools in relation to planning and the future. Along those lines, it might be noted that many American business schools hâve their curricula under re-examination and are moving to intégrale futures thinking into the Master of Business Administration programme. Professor Godet's work shouid make a spécial, positive contribution to that new move. The book shouid also find a substantial audience among individuals without planning responsibilities, vvho, out of curiosity or concern, seek a better understanding of the choices before us and the tools for exploring those choices, JOSEPH F. COATES

Coates &Jarratt, Inc. Washington, D.C.

Contents

1, FOR ANOTI-IER FUTURE 1 1 Dreams créait rcalky 1 ]. 1 The dangers of being reactive and the benefits of being pre-acîive and pro-active 1 1.2 Using intuition and reason for planning and action 2 1.3 From anticipation to action by means of appropriation 3 1.4 Contingency and change: ! ,001 keys to excellence 4 1.5

Towards a revival of planning 5

2 Scliools of prospective: modcls which endure in spiic offashion 6 2.1 The post-industrial school, bctwcen optimism and pessimism 6 2.2 The neo-Makhusian school ofthought 7 2.3 Long waves and crises as bearers ofhope 8 2.4 Bifurcations, determinism and creativity of chaos 10 2.5 Mutineers and mutants as die bearers of change 13 3 Schools of strategy: coniradictory fashions, complemcnlary modcls 15 3.1 The heuristic schooi: from cornmon sensé to the simplistic 16 3.2 The rationalise school: from success to misuse, pending a revival 19 3.3 Stop bacl American soaps and Japanese curios 20 4 Spreading the cultures of prospective and strategy 25 5

One logic in len chapicrs 26

FROM DETERMINISM TO DETERMINATION, FROM FORHCASTING TO 'STRATEGIC PROSPECTIVE' 29 1 Action-oriemed anti-fatalislic thinking 29 2 2.1 2.2 2.3 2.4

A critique offorecasting 32 Définitions 32 Inaccurate data and (instable models 34 Excessive quantification and extrapolation 34 The future explained in terms of the past 36

3 Usefulness and scope of prospective 38 3.1 Usefulness and credibility of prospective analysis 38 3.2 What ingrédients go into making an accurate forecast? 39 3.3 What makes a good expert? 39 3.4 The absence of neutrality of information and forecasts 40 4 4.1 4.2 4.3 4.4

Pluralism and complcmentarity 41 Problems and methods 41 Planning needs forecasts with scénarios 44 Proposais for solving the crisis in forecasting 45 Key concepts for a methodology 47

5 Practical advice for future thinking 47 5.1 Ask the right questions and mistrust convcntional wisdom 47 5.2 Think in the long term and imagine first what wilt not change 49 5.3 Use simple, adaptable methods to stimuiate thought and facilitate communication 50 5.4 Start with prospective workshops 51 5.5 Practical guideiines for prospective groups 53 T H E SCÉNARIOS METHOD

1

57

Origins and définitions 57

2 Types of scénarios and stratégies 60 3 Objectives of the scénarios method 63 3.1 Construction of the base 64 3.2 Building scénarios 67 4 4.1 4.2

Illustrations 74 Scénarios as tools of strategy and/or management 74 Successful utilisation of scénarios - the case of Elf 77

5

Anticipation and scénarios: myihs and rcalities 79

4. I D E N T I F Y I N G THE KEY VARIABLES: S T R U C T U R A L ANALYS I S 83 1

Origins and objectives of'structural analysis 83

2

Listing ail the variables 84

3

Location qf relationships wilhin the structural analysis matrix 85

4 Search for the key variables wilh the MICMAC® method 90 4.1 Direct and indirect classification (MICMAC®) 91 4.2 The MICMAC® principle: raising the power of the matrix 93 4.3 Comparison betwcen direct, indirect and potential classifications 96 4.4 The influence-dependence chart 99 5 Value ami limitations of structural analysis 101 5. U N D E R S T A N D Ï N G THE ACTORS' STRATEGIES: THE M A C T O R 0 METHOD 105 1 Constructing the actars' strategy table 108 2

Idenlifying the stratégie issues and associatcd objectives 111

3

Positioning each actor in relation to the stratégie objectives (signed position matrix) 113

4 Ranking the objectives for each actor (valued position matrix) and assessing the range of possible convergences and divergences 115 5 Evaluaiing the balance of power and formulating stratégie recommendations (valued matrix of position wilh power coefficients) 119 6 Key questions for the future 126 6.

REDUCING UNCERTAINTY: EXPERT C O N S E N S U S M E T H O D S 129 1 Scanning possible futures: morphological analysis 130 1.1 The field of possible futures^ a morphological space 130 1.2 Problems and limitations 132 1.3 The useful morphological subspace 132 2 An 2.1 2.2 2.3

inventory of expert-consensus methods 134 The Delphi method 134 Voting machines: Regnier's Abacus and the Consensor 137 Cross-impact methods 139

3 3.1 3.2

The SMIC method 142 The principle and the aim of SMIC 143 Relationships between hypothèses and scénarios: the search for cohérence 144

3.3 3.4

Results: hierarchy of scénarios and sensitivity analysis 146 Applicability and limitations 148

4 4.1 4.2 4.3

Case-studies 150 Air transport scénarios 151 Nuclear energy scénarios to the year 2000 158 Excerpts from the SMIC questionnaire 163

7. IDENTIFYIMG AND EVALUATING STRATEGIC OPTIONS 17 1 1 The décision process: identify, évaluais, choose and implantent 171 2 Relcvance irecs: frovi objectives to action 174 2.1 The qualitative hierarchy: aims, objectives, means and action 174 2.2 Flow chart notation and action évaluation 176 2.3 Usefulness and limitations 178 3

Classical metliodsfor options evaluating 178

4 4.1 4.2 4.3 4.4 4.5 4.6 4.7

Mukicriteria methods 182 Aid to decision-making 183 Listing possible actions 184 Analysis of the conséquences of actions 185 Defining criteria and evaluating actions in terms of thèse criteria 185 Définition ofpoticies and ranking actions 186 Multicriteria décisions in stratégie management 187 Sophisticated tools, limited applications 188

5 The melhod of comparative reverse ranking 189 5.1 The reverse ranking rules 190 5.2 The ranking of actions 190 6 6.1 6.2 6.3

The Electre II method 193 'Super-ranking'relationships 193 The ranking of actions 195 Conclusion 197

7 7.1 7.2

The Mithipol melhod 197 A simple operational tool 197 Case-study: launching a new product 198

8 . I N T E G R A T I N G SCENARIOS AND STRATEGV 2 0 7 1 Back to square one in stratégie management 208 1.1 Tvvo age-old concepts 208 ! .2 Two exemplary models: the Church and the armed forces 209 1.3 Two pioneers: Henri Fayol and Frederick Winslow Taylor 211 1.4 A prophet in the désert of management: Mary Parker Fotlett 213

2 The émergence and developmeni of stratégie management from 1920 to the présent 217 3 3.1 3.2

The inévitable encounter 220 The rise of prospective thinking 220 Questioning planning 221

4 An inicgraied approach: stratégie prospective 224 9. PEOPLE AND ORGANIZATIONS MAKE THE DIFFERENCE 22 9 1 Quality of leaders and witmitig teams 229 2

'Yes' to the plan and shared vision, 'no'10 ils pitfalls 232

3 Mctamorphosis ojstructures and behaviour 234 3.1 Adaptable or anticipatory structures? 235 3.2 Simple, lean structures 237 3.3 Stratégie centralisation and operational décentralisation 238 3.4 Stratégie conséquences of environmental changes 239 3.5 From an end to habits to a mental révolution 24t 10. T H E

DAWN

OF

THE

TWENTY-FIRST

CENTURY

247

1 Multiple nnceriainiies linked to the absence oj a regulator 248 2 2.1 2.2 3 3.1

A quasi-certainty: Europtimisw 251 From Ruropessimism to Europtimism 251 Europe 1993: the dream and the reality 252

Elcven likely trends 254 Démographie imbalances and South-North and Bast-North migratory flows 254 3.2 Threats to the physical environment and the négative iegacy of past growth 255 3.3 A lawless and turbulent international environment 258 3.4 Slow, irregular, unequa! and interdependent growth 259 3.5 New energy price rises: bénéficiai if they are graduai 259 3.6 The flood of new technologies: a new compétitive order 260 3.7 Deregulation associated with new international and régional régulation 260 3.8 Economie compétition on a global scale with states playing a key rôle 261 3.9 Décline in industrial jobs and rise of service industries 262 3.10 Crisis of the Welfare State 264 3.11 New forms of exclusion 265

4 4. i 4.2 4.3

Three major uncertainiies 266 Will the rate of diffusion of new technologies be as rapid as forecast? 266 Metamorphosis of work and cmployment 267 Evolution of lifesryles and social organization in developcd countries 268

5 The 5.1 5.2 5.3

lie of the land in the year 2000, in brie} 269 Probable trends 270 Major uncenainties 273 Possibie discontinuities (examples) 274

REFERENCES AND BIBLIOGRAPHY

275

1. For another future

1. Dreams create reality This book is, above ail, a manual of prospective and strategy; its main aim is to provide simple, operational tools for the conspirator of the future, which potentially means ail of us. What does this mean? How can we reconcile intuition and reason? How can we move from anticipation to action? What are the keys to excellence? What should we make of the successive fashions and models which dominate prospective and stratégie thinking? Thèse are the main thèmes which we shal! address in this introductory chapter, before outlining the book's format.

1.1.

T H E DANGERS OF BEING REACTIVE AND THE B E N E F I T S O F BEENG P R E - A C T I V E A N D P R O - A C T I V E

Unfortunately, anticipation is not widespread among managers. When ail is going well they can manage without it, and when things are going badly it is too late to see any further than the end of one's nose: one has to react, and quickly. However, reaction alone leads nowhere (remember Seneca's comment: 'There is no favourable wind for the man who knows not where he is going1). In other words, if there is no future direction the présent is empty of meaning. This emptiness is frightening; in the absence of a plan, too many managers thus find a constantly renewed outlet in the intoxication of action. Too many managers pedal like crazy 1

For anoiher future

from one stage to another, according to an itinerary imposed by the whim of circumstance. Absorbed in their efforts, with their no ses at the handlebars, they no longer hâve the time to think. Eventually, in full swing and on the verge of exhaustion, they iose their balance. Let us reiterate: action in the short-term reality has no meaning unless it takes place in the long-term context of a plan, because 'the future is the raison d'être of the présent'. Prospective is above ail an attitude of mind (anticipation and will) and a way of behaving (with imagination and hope) mobilized in order to ensure quality and control of our présent and future existence. Prospective reins ta tes désire as the productive force of the future. If it has no future direction the présent is empty of meaning. Thus, dreams do not oppose reality, they create it, and a pian animated by désire is the driving force for action.

1.2. USING INTUITION AND REASON FOR PLANNING AND ACTION

Whilst reflection must not paralyse action, pilotage with vision while keeping a look-out cannot replace strategy. It is because of past lack of anticipation that the présent is full of questions, which yesterday were insignificant but which today are urgent and which hâve to be settled now, to the détriment of what is important for the future. So it is time we stopped opposing pilotage with vision (intuition) and stratégie pilotage (reason). Both are necessary; it ail dépends on circumstances. A person cannot be reduced to a rational mind (the left brain); he is also driven by the emotional faculties (the right brain). Rational and heuristic schools of stratégie thought are only apparently in opposition; in fact they are complementary. If one or the other is denied, a person's behaviour becomes crippled, because he now has only a debilitated thought faculty. A sound initial reflection, imbued with relevance and cohérence, reinforces the efficacy of action and réaction in the face of events. In the rire of action the time for reflection is limited to a few moments of intuition and refïex thought. Intuition is the lightning flash which springs from the right brain through the synthesis of information and analysis stored in the left brain; hence the importance of initial reflection and of accumulating such reflection. The same applies to réflexes - they are better after intensive training. Being reactive should not be confused with being flexible. The first is improvised as an organization's appropriate response to external hazards.

Drcams creaie reaiily

The second is kept up as an intrinsic capacity of the organization to react and adapt to the environment without losing its direction. In other words, internai flexibility is the condition for external reactivity. The first dérives from physical condition, the second from réflexes.

1.3.

FROM ANTICIPATION TO ACTION BY ME AN S OF A P P R O P R I A T I O N

Intellectual and emotional appropriation is a compulsory stage if anticipation is to crystallize into effective action. Thus we discover the three components of the Greek triangle: prospective thought gives content to mobilization, maintains motivation (motives for action) and nourishes stratégie will. We can define thèse three components as 'Logos' (thought, rationality, discourse), 'Epithumia' (désire in ail its noble and not so noble aspects) and 'Erga' (action and realization). The marriage of passion and reason, of heart and mind, is the key to success in action and to individual growth (the body). Let us give the same message in colour: the blue of cold reason associated with the yellow of warm feeling produces the green of brilliant action. In the Greek triangle we find old debates which are still current. Thus men of thought complatn about not having a grasp on action, and men of action complain that they lack the time for refiection. One day we realized that the best ideas were those that people discovered by themselves. The phenomenon is well known: a good idea will move up the company hierarchy more easily if the boss believes it cornes from him; one should adopt this same reasoning with people one wishes to convince. Any thinking which is not appropriated by those concerned and which relates to change will hâve difficulty crystallizing into effective stratégie action. On the other hand, there must be a 'content' to be appropriated. Individual and collective motivation and mobilization are stronger and last longer when there is willingness to be drawn upon for a future project. The Greek triangle has its head at the base, in order to show that reason is not enough to ensure balanced action, and that passion is a necessary stage (Fig.l).

Far anotherfuture

Anticipation Prospective thought

Appropriation Collective mobilization

F I G . 1. The Greek triangle: prospective gives content and direction to collective mobilization. © M. Godet.

1.4. CONTINGENCY AND CHANGE: 1,001 KEYS TO EXCELLENCE According to the principle of contingency,1 there are no universal key factors of excellence which are valid in ail circumstances. For example, zéro stock and just-in-time hâve been profitable aims for some companieSj but also very expensive or even ruinous for others. Management theorists should recall that statistics teach us not to confuse corrélation with causality. It would appear that the authors of In Search of Excellence (Peters and Waterman, 1982) forgot this distinction. In their X-ray of successful United States companies in i 980, they noted that thèse companies had various points in cornmon which they immediately termed 'secrets of excellence'. Colinearity is not causality, however; they should hâve examined whether poor performance companies also demonstrated thèse same characteristics. In fact, the majority of companies cited in In Search of Excellence experienced difficultés during the 1980s. Should we conclude from this that the secrets of excellence were actually the secrets of failure? Certainly not - it is simply a matter of recognizing the principle of contingency and admitting that there are combinations of human, organizational, technical and financial factors which are effective in a particular given context but which wiil not be so in other contexts. Thèse combinations are multiple and change over tirne. 1. Contingency: that which couid not be, or which couid be orherwise; in other words, the opposite of necessity.

Dreams crcate reality

The only key factor of success whose permanence we can vouch for is preciseiy this principle of change, 'the source of diversity'. Biaise Pascal noted that 'boredom is born out of uniformity'. Variety is an indispensable stimulant for motivation - without it effectiveness is weakj with it anything is possible. Remember the experiments of Elton Mayo in the 1930s: by increasing and then decreasing tlie light intensity in the factories of Western Electric he increased output each time. A far-seeing manager is one who regularly introduces factors to break habits. In order to remain motivated a person needs to be permanently diverted and stimulated by challenges. Such challenges are more mobilizing if they are new, and more relevant if they form part of a cohérent trajectory. This is how we should understand apparently disjointed attempts to mobilize tlie intelligence of organizations: leadership by objectives, participative management; quality circîes, company plans, etc. Ali thèse attempts are marked by success and failure. In addition, reactivity and management in real time - which, as we hâve seen, it would be dangerous to make a religion of- can prove necessary and even miraculous in certain précise circumstances. A universal key to excellence does not exist - it must be permanentiy remodeiled according to environmentaî developments, counteracting the habits which fix organizations in the deep sleep of rigidity. What brought success yesterday could perhaps be the cause of failure tomorrow. In order to adapt to a changing worïd, we need to knovv how to change structures, behaviour and habits.

1 . 5 . TOWARDS A RE VI VAL O !•" PLANNING

Any exclusive practice quickly becomes abusive, so managers must be careful to avoid becoming addicts of reactiviry and real-time management. In plain language, we say yes to reactivity, but on the condition that it does not become an end in itself. Tactics, fhat is, contingent decision-making, should not take tlie place of strategy. From this point of View, the expression 'stratégie reactivity' is a source of misunderstanding, as normaily reactivity should be placed at tlie service of strategy. The same applies to so-cal!ed 'stratégie management', a beautiful pleonasm, since by définition the aim of management is to put tlie organization at the service of strategy. Jean Cocteau put it so well: 'Fashion is what goes out of fashion.' Soon reactivity will hâve had its hour and we shall rediscover, perhaps with new words, the virtues of stratégie planning - a skilful mix of voluntarism and flexibiiity. For this comeback to happen, tlie concept of

For another futurs

the plan wili first hâve to be definitively liberated from the bureaucratie corsets in which it has wrongly been laced.

2. Schools of prospective: models which endure in spite of fashion It is not by chance that prospective is less sensitive to the effects of fashion than strategy. As a matter of fact the horizon of prospective is usualiy long-term (defined as the horizon at which many things could hâve changed), whereas the horizon of strategy is generally shorter: fashion in clothing changes every year; fashion in housing does not. Although fashion is too ephemeral for prospective, it has nevertheless experienced différent schools of thought over récent décades. Eric Philipart (1986) has written an excellent synthesîs on this subject. Inspired by him we distinguish hère four schools: post-indu striai, neoMalthusian, long waves, and bifurcations and chaos.

2.1.

THE

POST-INDUSTRIAL

BHTWEEN

OPTIMISM

SCHOOL,

AND

PESSIMÏSM

This school of thought, born in the 1960s, a rime of rapid growth, is symbolized by the work of Daniel Bell (1976), The Corning of Postindustriai Society: A Veniure in Social Forecasting. The book's subtitle is not neutral; it conveys the idea that change, particularly technological change, can be accelerated and controlled by forecasting and planning. The methods exist-Erich Jantsch (1967) made a survey of them at OECD. This school of thought, on the whole liighly optimistic, had been championed several years earlier by a book by Herman Kahn and Anthony Wiener (1968) on the year 2000 (i.e. the équivalent of 2025 today). Kahn remained extremely optimistic until his death. His last book (Simon and Kahn, 1984) was The Resourceful Earth: A Response to Global 2000 (i.e. a response to a report to the United States président published in 1980, which was very pessimistic about future ecosystems). Alvin Toffier is probably part of this schooi of thought. A talented journalist, he received tremendous média exposure with Future Shack (1971), which was subsequently enhanced with his later books, particularly The Third Wave (1981). Thekey to success is simple: extrapolate technological change and announce the best sometimes, and often the

Sciwols of prospective: models ivhich endure in spite offashion

worst. As a good journalist, he knows that 'bad news is good news1. So in Future Shock he wrote: We are at the dawn of an international révolution which will overthrow parîiaments and houses of congrcss in the coming décade. But this raising of shields against the ravages of the ill-managed application of technoiogy could crystallize into a pathological form, like a futurephobic Fascism under whose régime men of science would replace Jews in the concentration camps. We shall make no further comment! None of this was very serious, but it did hâve the merit of making people think. Forecasting and planning's infatuation with controlling the future came to an end with the oil shocks and the économie crisis. However, several years later, at the end of the 1970s, the champions of the third industrial révolution, which it was believed would ensue thanks to new information technologies, biotechnologies and new materials, rekindled the fiame of the post-industrial society. Daniel Bell kept silent, and technological determinism was reborn out of his ashes. It was a vehicle for unfounded clichés about the society of tomorrow, such as the qualifications required for future professions. Tomorrow's jobs would be in the services, linked to the major trends of modem societies (ageing, leisure, security), which meant, in particular, care in the home, house maintenance, gardening - ail jobs where compétition and demand for quality would impose high standards of professionalism, which should not, however, be confused with high qualifications.

2.2.

T H E NEO-MALTHUSIAN

SCHOOL OF THOUGHT

This school of tliought was characterized by the Club of Rome, founded in 1968 by industrialists, académies and high-level civil servants. In a way this school of thought is also the heir of the 1960s wave of rationalism. It rested on the technique of System dynamics, developed at MIT by jay Forrester. The first report of the Club of Rome by Meadows et al. (1972), The Limits w Growth, had an explosive impact. Exponential growth could no longer continue, resources would run out and the worid would collapse under démographie pressure. The report did not really pose the right questions, but it did hâve the merit of provoking a great debate, although it was quickly eclipsed by the économie crisis. OECD's 'Interfuturs' report (Interfuturs, 1979), produced under Jacques Lesourne in 1978, showed that there were not really any physicaî limits to growthj but more particularly problems of régulation on the

For anoiher future

global scale in an interdependent vvorld where, according to Daniel Bell's famous phrase: 'Governments hâve become too small for the big problems, and too big for the small problems.' Ecology was part of thèse big problems, and the Giobal 2000 report in 1980 began with the following terrible prédiction: 'If current trends continue, the world in the year 2000 will be more polluted, more congested . . .'. In an interview in 1982 Gerald Barney, the author of the report, revealed to us that this first phrase had been censored - that is, remodelled; the original version was: 'If current policies continue . . .*, which is not at ail the same thing. The diagnosis was much less fatalistic but also more accusatory of those who govern us. So the Interfuturs author s were right - it was régulation which was lacking. Since the end of the 1970s there hâve been two other schools of thought: long wave theory and bifurcation theory.

2 . 3 . L O N G WAVES A N D C R I S E S A S B E A R E R S O F H O P E

The theory of long waves re-established Kondratiev's analyses, which explained successive waves of growth and recession by waves of innovation. Each phase of the growth-recession cycle would last for about twenty-ftve years. Thus, after the high growth of 1950-74, there would be a trail across the désert undl the 1990s. Then would begin the era of a fifth Kondratiev cycle, supported by the technological révolution which had been latent since the 1970s. This theory of technological long waves spawned the neo-Schumpeterian gurus, wel! represented by the works of Christopher Freeman (1974) and Scherer (1986), or in France the so-called régulation school. Thèse works generally emphasized socio-organizational obstacles, that is, the rigidities opposing transformation of production structures and curbing the structural changes which the new technologies would necessitate. Part of my own earlier work to a certain extent descended from this line - but not totally, for I emphasized above ail the actors (in crisis) facing Systems (undergoing change). From this perspective, crisis is perceived both as the conséquence of rigidities and as the main lever to overcome tliem, changing as it does our habits, structures, and behaviour. Crises are bearers of hope and capitalism is a perpétuai hurricane of créative destruction (to quote Schumpeter). The titles of my books over the years - Demain les crises (Tomorrow Crises) (1980) ; Crises Are Opportunities (1985) and (with Jacques Lesourne) La fin des habitudes (The End of Habits) (1985) - provide a good résumé of my interprétation of crises and change.

Schools ojprospective: models îvhich endure in spite ojfashion

But there is a fondamental différence between this work and that of the supporters of long waves and some of their gurus. I hâve always challenged technologica! determinism, which in our modem societies has taken over from the religious determinism of the past (witii the technological fairy replacing God). There is no fatalism. There are, on one hand, forces of change and, on the other, forces of inertia. Often thèse are in opposition and we speak of résistance to change or (economically, socially) non-viable change, depending on which is in favour. Structures and behaviour are characterized by strong inertia, which curbs adaptation and évolution and then becomes rigidity if change (whether desired or not) has not been prepared for well in advance. If long waves exist, they are socio-organizational rather than technological, i.e. they are linked to the non-linear rhythms of societies and people. Unlike people, society shows no regularity between cycles of sleep and intense activity. Random external and internai circumstances can precipitate évolution in an unexpected direction and, for example, give power to certain actors who, armed solely with their will and their plans, manage to overturn the course of events and cause bifurcations. The collapse of the communist régimes in the countries of Eastern Europe provides a good example of bifurcation - it was unforeseen and nothing enabled us to forecasl it, even though the forces which ied to the splitting up of the empire had been spotted a long time ago (I refer to the work by Hélène Carrère d'Encausse (1979), L'empire éclate). It was assumed that thèse forces were oniy at work over the very long term and scénarios of violent disturbances and repression were imagined, but never scénarios of almost calm évolution, with a forced march towards pluralist democracy and the market economy. A short time ago it was supposed that German réunification woutd inevitably be accompanied by the neutralization of Germany and the Finlandization of Western Europe . . . Today's reality has overtaken yesterday's fiction. It is Eastern Europe that is westernizing and the former German Démocratie Republic which is joining NATO. We rejoice at ail this. However, the new equilibrium is unstable and could lead to other dangerous bifurcations, particularly in Russia and the new republics, where internai difficulties are weakening the reformist forces. We should not forget the lesson of history: external war often served as an outlet for internai contradictions. The prospecrivist will bear in mind that one shouid remain modest in the face of the future - that nothing is ever hopeless or established, and that probably part of what is still fiction today will be reality tomorrow. Which part, where and how? We shall not reply recklessly,

For another future

but say simply that one must remain ready to face any eventuality, vvhich means, for example, not letting down one's guard on the subject of Europe an defence.

2.4.

BIFURCATIONS, DETERMINISM AND C R E A T I V I T Y OF CHAOS

At the time vvhen thèse upheavals are occurring (still contingencies), the bifurcation and chaos théories from the hard sciences (mathematicSj physics, biology) are encroaching into the social sciences. It ait began at the end of the Iast century with the work of the mathematician Poincaré (1889-1908). He showed that certain Systems of parametric équations had no stable solution, but an infinité number of unstable solutions, which could go from one extrême to another as a fonction of the parameters and of time. This is Poincaré's 'mathematical indeterminism'. Thus, there exist sets of équations which are sensitive to the initial conditions vvhose solutions diverge in a chaotic manner. It is also to Poincaré that we owe the notion of the 'diagram of bifurcations'. We note in passing that there is a double uncertainty hère: uncertainty over the measurement of the initial conditions and uncertainty over the conséquences of this lack of précision for the calculation of the solutions. In such Systems, determinism would be hidden, would hâve ail the appearance of chance, and woutd in any case remain indéterminable (if only because of the lack of précision of die original measurement). Much later thèse ideas were taken up again by Lorenz (1963) in simulating weather-forecasting models. He showed that a slight initial fluctuation could reverberate in an exponential and chaotic way over time. This idea was made famous by the image of the 'butterfly effect', according to which the simple beating of a butterfly's vving in one part of the world could in an extrême case cause a storm in another part some weeks or months later. Thus we once again hâve die idea of a 'deterministic chaos'. In the 1970s Uya Prigogine and Isabelle Stengers (1979) used the théories and experiments of thermodynamics and physical chemistry to go further and develop the concepts of 'order through fluctuation' and 'créative chaos' (Prigogine, 1990). They showed mat, 'far from the initial conditions of equilibrium', there appeared bifurcations leading to other equilibria. Near dièse critical points, slight internai or externat fluctuations in the system can be décisive in determining whether movement is towards one branch of évolution or another. Transposed to social Systems, thèse slight fluctuations could be, for example, chance 10

Schools of prospective: models which endure in spite offashion

disturbances or individua! actions. At thèse critical points 'fluctuations are dangerous', whereas otherwise they could remain insignificant. This reading of social Systems based on analysis of the évolution of physical Systems is interesting for prospective in that it reconciles determinism on the one hand (the System has a history which conditions its future possible trajectories, and leads- to one diagram of bifurcations rather than another), and freedom on the other (in the bifurcation zone insignificant actions, caused by chance or by free will, can set off major upheavals). Outside thèse critical points, fluctuations are not events. In thèse new interprétations of évolution, chance, necessity and free will, alternately and together, play a determining rôle. Other scientists from biology, such as Henri Atlan (1979), participated in this movement, implicitly taking up once again Darwin's principle of natural sélection, adaptation and diversity (1859). Certain chance mutations (fluctuations) vvithin living Systems constitute events. Events can also resuit from fluctuations in the environrnent and cause new adaptations — 'complexity through noise' — that is, a superior variant of the organism/superior mode of diversity in die System. Chance is the driving force of évolution, and noise becomes information. In short, this is 'chance as organizer'. Is it possible to adapt thèse descriptive théories of the évolution of physical or biological Systems to explain die transformation of social Systems? With die current state of knowledge we hâve to answer in the négative, which is just as well. It is not that we are veiling our faces in horror and retreating into some kind of obscurantisrn (rejecting science and its advances). It is simply that the rea! world is much too complex for us to hope mat one day we will be able to find an équation to explain its hidden determinism. Even if we could, the inhérent uncertainty over any measurement, especially socia! measurements, would mean diat, at least in our minds, the range of possible futures would remain wide open; rightly or wrongly, this is our mental image of the famous 'diagram of bifurcations'. So we hâve to act as if chaotic determinism did not exist and take action in order to move in die direction of those evolutionary branches which seem désirable to us. We prefer Pascale's bet and die mydi of Sisyphus tirelessly pushing his rock to the fatality of determinism. The expression 'chaotic determinism' seems doubly unacceptable to us, since it renders human will powerless in die face of the dual games of necessity and chance. Bifurcation and chaos theory above ail demonstrate, ex posi, that change in structure corresponds to one solution of a System of differentiai équations - for a given value of die environmental parameters. However, 11

For anatlier future

ex ante the question remains., vvhat will be the value of thèse environmental parameters and their eventual évolution? This problem has to be overcome in order to détermine whether the future which will materialize out of the range of possible futures is a rupture, a disaster, or not. This is precisely die aim of prospective - to détermine what could be the values of the environmental parameters, taking account of forces at vvork and actors' plans, i.e. what are the possible, reaiizable and désirable scénarios regardless of any offence thèse might cause. Bifurcation and chaos theory relaunched 'La querelle du déterminisme' (the quarrei about determinism), to quote the title of a collective work published in 1990 (Asterdamski et ai.). The resolute rejection of determinism in the social sciences should not, however, lead to a restraint on research in fondamental science. René Thom, the author of catastrophe theory,- seems to be isolated in this quarrei. Yethe is right to say that: 'Determinism in science is not a given, it is a conquest, The enthusiasts of chance are apostles of désertion' (see his article entitled 'Halte au hasard, silence au bruit' (Let Us Hait Chance and Silence Noise) (Thom, 1990.) Many European readers did not corne across this debate, which is now so fashionable, until the highly publicized success of the book on chaos theory by American journalist James Gleick (1987). The bestselling management authors sensed that it was a gold mine, and in 1987 Tom Peters published Thriving on Chaos. In 1986 Eric Philipart wrote: 'In the USA, the utilization of thèse concepts seems still to be timited to the adoption of new metaphors (without attempting to demonstrate the validity of this procédure) as an easy way of giving discussions otherwise lacking in great originality the appearance of novelty.' In Europe chaos theory is favoured by the média. It is true that the vocabuiary iends itself to journalistic emphasis and whets the appetite with words which make us dream, such as 'strange attractors', 'fractals', 'bifurcation cascades', 'the butterfly effect', etc or which inspire fear, such as 'noise', 'disorder', 'catastrophe', 'chaos'. 2, The terni 'catastrophe theory' has become popular, but it is often misundcrstood, bcciuisc of the vcry strong impact on the imagination of the word 'catastrophe'. For Thom, a catastrophe is the disappearance of onc cquilibrium and the establishment of a consécutive one with a continued modification of potential. Simple cxamplcs are provided by the paasing from a liquid statc îo a solid or gascons state, or the metamorphosis of the chrysalis into a bunerfly (it is the same but it is complctcly différent). On this subject we share îhe views of the mathematician Ivar Eckciand (1984): 'Catastrophe theory has no vocation to be normative or even to predict . . . it docs not announce anything, not even catastrophes . . . the success of this theory rests partly on an initiai misunderstanding caused by the magie of words.' 12

Sc/mols of prospective: niodels whiclt endure in spile offashioti

It is alvvays good to enrich one's vocabulary, especially when thèse concepts écho well-established mathematical conventions or physical mechanisms. The scientists who write on thèse subjects generally respect the potential and the limiis of thèse concepts. They oniy venture to transpose them into the social domain as questions, and vvith extrême caution. They will nevertheless be reproached for not always giving the same meaning to the same words - giving rise to misunderstandings which cause epic quarrels between specialists about chance, determinism, etc. Unfortunately, in the field of social sciences the 'fluctuation' in the movement of ideas brought about by bifurcation and chaos theory lias only added to the disorder and confusion in 'houses' where the housekeeping of ideas leaves much to be desired. We are stitl a long vvay from the 'science of complexity' hailed by some experts (cf. Amara, 1984). Meanwhile it is primarily complication that we are confronting. In fact, complexité' has become a great cooking pot into which we haphazardly tlirow ail the concepts we find on the table of advanced ideas (epistemology, information theory, thermodynamics, dissipative structures, bifurcations and of course chaos, not to mention self-organization). To give the resulting soup a réputation of good cuisine we invoke the names of ail the saints (scholars, intellectuals) who protect thèse concepts. But we forget that they are not ail from the same church. Thus, the so-called 'science of complexity' is rather like a new religion which claimed to be superior because it borrowed from Islam, Christianity, Buddhism, etc. It is an idea than we should ponder over no longer than we would over Espéranto. Representational models do not escape the vagaries of fashion, and fashion always relies on models. This is as true for prospective as it is for strategy and management - except that prospective has a greater propensity for models, and strategy has a greater sensitivity to fashion.

2 . 5 . M U T I N E ERS A N D M U T A N T S A S T H E BEARERS OF CHANGE

Where should we place prospective in relation to thèse schools of thought? It has to be said that prospective constitutes an unclassifiable mosaic which borrows the most illuminating aspects of various readings, From religions determinism prospective retains the fact mat the question of why (the clockmaker) is not raised by the answers to the question of how (the ciock). Science is advancing but the idea of God does not recède; it continues to assert itself as a defence against the absurd. 13

For another future

The theory of évolution has brought to prospective tlie principles of natural sélection and adaptation, which hâve lost none of their modernity, for both individuals and for organizations facing international compétition, as Hubert Landierhas shown clearly (1987). The mechanisms exist; it is up to us to make use of tliem for our own beneftt, to adapt to and to transform our environment. Actors are not disarmed in the face of Systems: they can act either to drive or to restrain évolution. The inertias of structures, behaviour and habits are such that often it is necessary to destroy former organizations in order to create new ones. From chaotic determinism prospective wili retain bifurcations and the possibility that certain mutineers or mutants may be bearers of change: 'Not any individual, idea or behaviour, but the "dangerous" ones, that are ab!e to use for their own benefit those non-linear relations which would ensure tlie stability of the former average state' (Prîgogine and Stengers, 1979). How are we to recognize bifurcation points? 'Which events, which innovations will remain without conséquence, which otliers are capable of affecting the whole System and determining irreversibly tlie choice of évolution; where are the zones of choice, and the zones of stability?' Thèse are the questions which make up the daiiy menu of prospective. When we identify tlie range of possible futures are we not also recognizing the diagram of bifurcations? Are not the parameters of bifurcations also the key variables of prospective analysis? Since determinism is indéterminable, we must act as if nothing had been set in motion, as if only tlie rébellion of free witl could overturn tlie tyranny of chance and necessity. For sorae years we hâve noticed that scientific work has been converging on tlie concept of self-organization. What are we to make of this new fashion of self-organization? Is its success due to the iact that it is both suggestive and fuzzy? It is as much an écho of Fourierism or workers' struggles as it is of cybernetics (Von Foerster) or Prigogine's 'order through fluctuation'. Now it is penetrating économie science (cf. Lesourne, 1990). In its new acceptance it forms part of the gataxy of concepts mentioned above. What is it ail about? For Jean-Pierre Dupuy (1982), 'The selforganizing process seems to be a resuit of interaction between an autonomous structure and chance events which are extraneous to it. Because it thus allows adaptation to the new and to production again selforganization achieves, at least localiy, a reversai of time's arrow.' This 'reversai of time's arrow' is no more than what prospective means by 'The future is the raison d'être of the présent'. In other words, 14

Schools of strategy: contradictoty fashions, complemeiitary modcls

we can say that désire as the productive force of die future is also die main driving force of self-organization.

3. Schools of strategy: contradictory fashions> complementary models Fashion has a short memory, Many ideas on organizational management and strategy are worthy of die famous comment recalled by Georges Yves Kervern (1986): 'Good ideas and new ideas, unfortunately, are not die same thing.' In fact, in stratégie circles new things are often not interesting and interesting things are often not new, as we can see if we look back to the sources of military stratégie diinking (Clausewitz, General Beaufre), and to the founders of managerial thinking at the beginning of this century (Fayol, Taylor, Mary Parker Follett). On this subject, Michel Crozier remarks that, despite its inadequacy, 'fashion offers new ways of responding to die new problems' which regularly appear and as a resuit of which former problems go out of fashion. Of course, thèse new problems are pardy generated by die excesses of previous fashions. Thus, fashion is renewed and destroyed by the successive waves of rigour and imagination which by turns are elevated as dogma. It has to be said that this succession is onty apparent to specialists, and diat consumers dress themselves from one source and anodier in strategy's ready-to-wear boutiques, depending on their history, thetr resources and die information available to them. Some su'Sl choose goaloriented planning or strategically segmented profit centres (linking die product with die market), others go for company plans togemer with quality circles. In order to be like die Japanese, some even go so far as to turn the company into a place of regimentation and psychological conditioning. Heads of companies believe diemselves to be messiahs and hammer out their 'ten commandments'. The reality is simply a mixture of ail this, confused radier dian intelligent, where we juxtapose the pièces of die puzzle but do not really manage to put them together. The right and left brain form a whole which cannot be dissociated. Warm values (endiusiasm, charisma, désire, plans, wiil) and cold values (reason, analysis, calculation, forecasting) are not in opposition but complément and mutually reinforce each other. 15

For another future

This fact resurfaces throughout the literature on management. We can quote the excellent Riboud (1987) report Modernisation, mode d'emploi (Modernization: Directions for Use): Today there are comparées performing wcli in ail branches of industry, even the most vulnérable. It is firstîy a question of peopie with indisputable profcssional credibility and proven charisma. Today, motivation of peopie is the sine qua non condition for effectiveness and profitability. . . . Peopie are motived if their deep-seatcd aspirations are respected. To do this, managers must listers to them and, through negotiated agreements, arrange with them proposed objectives and positive sanctions. They should neglect neither the aspiration to be nor the aspiration to hâve, This complementarity does not prevent division, or ratlier a dialogue of the deaf, between, on one hand, supporters of a rationalist approach (économie analysis, operational research, segmentation, portfolio analysiSj planning) and, on the other, supporters of an approach fittingly called heuristic by Marie-José Avenier (1985), because one of its fundamental principles is to act and then to learn from action. We shall stick to thèse terms, which are very close to those suggested by odner authors, such as Thomas Durand (1985) and Alain Charles Martinet (1983).

3.1.

T H E H E U R I S T I C S C H O O L : FROM C O M M ON S E N S ! ; T O T H E S I M P L IS T I C

The heuristic approach excited Worldwide interest with the success of the book by Peters and Waterman (1982), In Search ojExcellence, and in France with the first book by Hervé Serieyx (1982), Mobiliser l'intelligence de l'entreprise (Mobilizing Corporate Intelligence), and his second book., vvritten with Georges Archier (1984), L'entreprise du 3ème type (The Company of the Third Kind). Thèse books insist, rîghtly, on the importance of the human factor - which managers must know how to motivate, mobilize, make responsible - the indispensability of listening to the customer, and the necessity of (perceived) quality of products and services. For supporters of the heuristic approach, thought should not be allowed to paralyse action. They claim that the clever methods of the rationalist approach hâve often led to failure. They prefer to proceed by trial and error, by an incrémental approach, developing what works and abandoning what doesn't. So far so good, but care should be taken not to throw the baby out with the bathwater. 16

Schooh ofstraiegy: contradictory fashians, compkineiuary modeh

The defkiencies of stratégie pilotage musr not lead to pilotage by sight. There must be a meaning to the mobilization of people; it is not an end in itself, flexibility and pi!oting by sight lead nowhere. Moreover, rationalists were quickly able to poke fun at their detractors. In fact, in autumn 1984 a celebrated Business Week article tempered initial enthusiasm for the secrets of excellence: a significant proportion of the companies presented by Peters and Waterman in 1982 as successful examples of collective mobilization were in difficulty. Hère was proof that the gymnastics of collective mobilization are not sufficient to ensure flexibility and performance in ail circumstances: in addition, anticipation of change and planning of strategy and innovation are necessary. A sound initial reflection, imbued witli relevance and cohérence, reinforces the efficacy of action and reaction in the face of events. In the lire of action the time for reflection is limited to a fevv moments of intuition and reflex thought. Intuition is the lightning flash which springs from the right brain through synthesis of information and analysis stored in the left brain; hence the importance of initial reflection and of accumulating such reflection. The same applies to réflexes - they are better after intensive training. Accused of gross simplification, some champions of the heuristic approach reply that their message is still too complex. Carried away by their 'passion for excellence', they turn it into a religion and their book becomes a catechism where company heads are transformed into mythical heroes playing the rôle of apostles. They thus forget that enthusiastic mobilization around charismatic leaders is not an end in itself, and that flexibility cannot replace strategy. The Gospel according to Saint Mac had the fate it deserved at the pen of Georges Yves Kervern (1986) (seebox). The Gospel grew by 610 pages in one fell swoop with Tom Peters' book, Thriving on Chaos (1987). The author does not beat about the bush. This book, he writes, throws into question ail the concepts of management. Today, flexibility and a passion for change must replace our inclination for mass production, destined for mass markets and based on the possibiiity (iong past) of making relativcly reliable forecasts. . . . tomorrow's champions will move ahead of chaos. This is total deiirium - ail the more paradoxical as the success of this kind of book is precisely the resuit of the effectiveness of a System of mass production and distribution (sales in garage and airport kiosks). In the form of pilotage by sight the heuristic approach is quite insuffkient and may lead to worse just as easily as better. 17

For anoiher future

The Gospel according to Saint Mac, by G. Y. Kervern {Annales des Mines, Gérer et Comprendre, March 1986) 'Good jdeas and new ideas, unfortunately, are not the same thing'. This cruel phrase, which certain professors amuse themselves by writing on their students' work, would not be out of place describïng the writings of the Apostlcs of Excellence, the celebrated Thomas J. Peters and Robert W. Waterman, both educated at the school of McKinsey and adtnired advocates of what could be called 'The Gospel according to Saint Mac', which has sold in its millions. The Gospel runs to 806 pages of which 306 pages for the Old Testament {In Search of Excellence, 1982) and 446 pages for the New Testament (A Passion for Excellence, 1985). Among those ideas which are good but not new, we find ourselves confronted with 'what every chief executive should know': be a good boss, take care of your clients, look after your staff, be vigilant about quality, simplify as much as possible, do what you know how to do, decentralize, and use small motivated groups for innovation. Thèse are the much-vaunted wonders of excellence which everybody is talking about today as if this were a historié discovery in the annals of management. We can cenainly rejoice that thèse basic truths hâve been esteemed so highiy. We may déplore the fact that they are sometimes ignored by certain chief executives. We can be glad that our US competitors hâve a tendency to ignore them. Wc may tremble at the idea that the japanese are putting them scrupuiousty into practice. But do we need to pay S 12,000 for each conférence devoted to the articulation of thèse holy princtples? Among those ideas which are new but not good, at least two are currentiy enjoying a great succès d'estime, in California at any rate: • mistrust of human reason; • mistrust of respect for rules. Thèse two ideas hâve culminated in rhe paradigm of the 'skunk'. For the authors of A Passion for Excellence a skunk today refers to individuais who pursue a highiy créative activity on the margins of a company'.s formai Systems. There are two rules in the skunk catechism: • act first, ask questions Iater; • 'cheating' is OK, and more efficient. The skunk thus has short ideas. His drug is action, his slogan KISS (FCeep It Simple and Stupid). The word 'scupid' is very important since it accurately sums up the mistrust of the rational model which forms a basis for that 'coming révolution', Let's pray that this KISS, apologist for the rough and ready and the stupid, does not turn out to be the KISS of Death for Science and Management.

Schools ofstrategy: contradictoiy fashiom, compîetnentary models

3.2.

T H E R A T I O N A L I S T SCHOOL: FROM SUCCESS TO M I S U S E , P E N D I N G A REVIVAL

As Marie-José Avenier has quite correctly written (1985): 'If there is a dearth of literature on the operating modes of pilotage, we are, on the otlier hand, witnessing a véritable burgeoning of practica! tools.' Thèse planning tools are well-known: life cycle, the expérience effect, stratégie segmentation, portfolio analysis, muiticriteria choice, etc. The panoply has recently been enriched by compétitive analysis, value chains and compétence trees, and it's not finished yet, since scénarios from elsewhere (prospective) are now disembarking on PSanet Strategy. Thèse rational methods found great success in the 1960s and 1980s. They had the great merit of imposing a common language and relevant modes of reasoning which allowed générations of managers to be trained in strategy. Since then the concepts of segmentation, of milch cows and of stars hâve been part of ail managers' vocabularies. The champions of the rational methods often went too far5 however. They were tempted to présent one scénario or another as the universal key to stratégie analysis. Reality rebels against panaceas and always contradicts theory with solid counterexampîes. Thus, for example, according to the expérience effect cited by the Boston Consulting Group, companies having the strongest market share should be the most compétitive. How then can we explain the better performance of smaller companies with a weak market share? The answer is probably to be found in the area of innovation and in the diversification which it permits. Is not the accumulation of expérience with a given technology and mode of production a source of rigidity, a brake on évolution? The effects of volume, quality and innovation ail, to a greater or lesser extent, play a rôle in profitability, as the authors of the PIMS method (databank on the performance of companies which can be used for simulation) hâve shown (cf. Buigues, 1985, and Thiétard, 1984). Each tool of the rationalist approach is partly true but also partly false. The degree to which it will be one or the otlier dépends on many other factors, internai and external to the company, which cause the latter to be understood as an élément of a System, of a network, or even of a network of actors and System. In tliis context, Georges Yves Kervern and Jean Pierre Ponssard hâve noted (1990): 'The search for rationality therefore becomes the search for a language adapted to an imperfect state of knowledge which it is nevertheless in one's interest to share in order to structure collective action.' In otlier words, the rational 19

For anotlier future

approach must pollinate the heuristic approach to give action the intelligence, analysis and thought which it lacks. In this way we find the complementarity between thèse two currents, mentioned above, which it is useless to oppose. Stratégie pilotage needs the compass, instruments and maps of the rationalist school of thought.

3.3. S T O P B A D AMERICAN S O A P S AND J A P A N E S E C U R I O S

The rationalist and heuristic schools hâve a tendency to follow fashion and to forget the expérience of the past. Thus the SWOT approach of looking at external strengths, weaknesses (human, technical, financial), external opportunities and threats, which vve owe largely to the pioneering work of Igor Ansoff (1965), has appeared out-of-date to some people. It's ail very weil to speak of value chains, compétitive environmental chains, stratégie positioning, or the mobilization of people; to start vvith, one should really know one's products, one's markets, one's costs and one's human, technical and financial resources. The stratégie diagnosis's prescription dépends above ail on the conclusions of this classic check-up. This is how celebrated books speak of business strategy, without saying a word about financial or human aspects. Speaking frankly, after many and careful readings, I am convinced that most bestsellers in the field of management are the équivalent of American TV sériais. Thèse 'products', usually 'made in the USA', repeat the obvious, which is applicable to everybody, and from gênerai conclusions draw simplistic examples which are easy to understand without any great effort. They are sold in airports like magazines. Letus hope that thèse intellectual sandwiches don't spoil die appeu'te of those just starting out. For the rest, they are not dangerous, since they contain virtually nothing worth remembering. Most commonly in thèse vvorks, usuaily several hundred pages long, an avalanche of uninteresting maxims of the type 'the boss puts his desk in the middle of the typing pool' (Tom Peters and Nancy Austin) substitutes for démonstration. But, you will say, this définitive comment is particularly applicable to the heuristic school. Alas, not really! In factj certain stars of the rational school, like Michael Porter, only partly escape criticism. It is not the content which is wanting, it is the spirit of synthesis - and the reader finds him- or herself drowning in a Dde of analytical détails. Thus, for example, in the interplay of actors in the compétitive environment (Porter, 1980), ittakes some time to realize that the author is repeating the same thing over and over again by 20

Schools ofstrategy: amlradîcwry fasltiom, compkmeniary models

considering, each time, each one of the actors in relation to ail the others. Reasoning is diluted by the encyclopédie volume of the writing. The prudent and efficient reader contents him- or herself with the summaries. At the same time, in-depth criticism is becoming rare. One needs to go right to the end and keep a clear head to notice, for instance, that the firm is scarcely ever approached from the angle of its financial reality and human identity! Mischievous longues are proposing a remedy for the tedium of American soaps on management: wait for the next épisode (book), which will hâve the same defects but the advantage of starting with a short résumé of the previous épisode. The latest vogue to date consists of marrying management with chaos (theory) (Peters, 1987). We hâve already discussed what we think of this unconsummated marriage. As with a firework display, we must end this introduction with the finale announced right at the beginning. This involves one of the most famous japanese curios, a 'crue lie' which made directors tremble, shook politicians and filled the front page of magazines: everyone gulped it down and it continues to circulate Worldwide. Let's tell this incredible taie. Everything started with those terrible comments attributed to the président of Matsushita: 'We are going to win and the industrialized Western countries will lose. You can no longer do much about it because you carry the seeds of your defeat within yourseives. Your organizations are Taylorist, but worse stiil, so are your heads . . . whereas we are post-Taylorists . . ." Widely quoted since the middle of the 1980s, thèse comments provided a saluîary electric shock to the brain: Japan's advance could be explained by its management model, which must first be understood in order then to draw inspiration from it. However, it has been known since 1987 that this rext is a 'true lie' - its author, Serieyx, revealed it in his préface to a book by Isabelle Orgogozo (1987) and confirmed it in 1989 in a new book entitled !^e zéro mépris (Zéro Scorned). 3. This œxt w;ss citet! by G. Arehier and M. Serieyx in L'entreprise du 3bma type, with the subtitlc 'A Japanese Présidents Speech'; then, wkhout quite knowing how, rumour attribuiecî it to îhe président of Matsushita. For the sake oi'compleîcness, the rest of the test is given herc. The ideas and style of Hervé Serieyx arc casily discernible. 'You arc totally persuadée! tluit you makc your businesscs function wcll by distinguishing bosses on one skie, executives on the other, îhinkers on one side, those who sit right on the other. For you, management is the art of making the bosses' icteas pass convenientiy into the hands of ihe workers. We are post-Taylorists, we know that business luis become so complicated, so difficulté and the survival of a firm so problematic in an increasingly dangerous, unexpected and compétitive environment, that a firm must mobilize îhe intelligence of everyone every day to hâve a chance of making it. For us, management is precisely the art of mobilizing and stacking up ail this total intelligence in the service of the company project.' 21

For anoiher future

Thus the Japanese system of management would hâve become a real fashion based on a false mode]. It would be a kind of rumour, but it is well known that, in order to circulate, rumours need to écho realities. We know that Japanese courtesy is falsely modest and often hides suppressed arrogance. The text was plausible. What a shame! For once a Japanese seemed to be telling us a few home truths. The story would hâve stopped hère if I hadn't been invited by the Management Centre of Europe in June 1990 to speak to 300 managers from around the world who had corne to Paris for the sole purpose of listening to eminent people such as Henry Kissinger make pronouncements on the prizes to be won and lost today. Among the speakers, I had the opportunity to listen to R. T. Pascale (Pascale and Athos, 1984), Harvard professor, recognized as one of the best specialists on Japan and its management model. Surprise was followed by inward shock - in the middle of his speech, he let loose, in English, the infamous quotauon attributed to the président of Matsushita. What was I to think? What was to be done - but sound the alarm and check what was going on? By questioning Pascale's assistant, I discovered that he had only been using the quotation for two or three years, while his last interview with the président of Matsushita had taken place in 1982. Persevering, I pushed curiosity so far as to (discreetly) challenge R. T. Pascale himself. Visibly embarrassed - to his crédit - he told me that in fact 'he had one day read this ten-line text! (the English version is identical to the French). He added, in connection with something else, that he 'had met the président of Matsushita in 1982 and that the latter, who was very old, spoke English almost inaudibly'. However, what he 'believed he understood from die président corresponded exactly to what he had later read as one of his pronouncements'. Perhaps Joan of Arc bas rivais across the Atlantic! To clinch die matter I neverthefess reported ail diis to Serieyx, who laughed hearuly and swore to me that we were definitely dealing widi a 'true lie' (and not 'a false true-lie'), and that he had not translated some obscure United States journalist. Serieyx specified that the quotauon had been in circulation and published since 1978, following a trip to Japan, in a business review. Will this déniai succeed in snuffing out the rumour? Rumour is already playing for time by putting the déniai in doubt, Who knows Hervé Serieyx in die United States? Not, in any case, Professor Pascale's assistant, who happens to be French-born: 'Never heard of hinV was her comment. Americans are ignorant of work not written in English, i.e. the greater part of the world's writing . . . This true-fatse text is only one of a number of Japanese 22

Schooh of strategy: contradictoryJashio7tst cotnpleiiiematy tiiodels

pronouncements used as références or bogies.*1 We are dealing with a real management fashion based on a false model. What share does reality hâve in the images we receive of Japanese management? In order to explain the formidable competitiveness of Japanese firms, we should first cite the will to work to escape the privations of the 1950s, the memory of which is still fresh in people's minds. Japan has no other resources than peopie and their labour. 'Just-iri-time' production methods were developed there because of the acute lack of space and the costof maintaining physical stocks caused by the very high price of land (four-fifths of japan comprises uninhabited wooded mountains). This is why Japanese factories are so compact, which requires great efforts in organizing production but ultimately improves productivity. This model should not be engraved on tablets of stone; Koreans are often more compétitive than the Japanese, with huge factories where zéro stock is not the rule. Concepts of productivity and quality hâve been imported from the United States, where, in the 1950s, productivity was the highest in the world. Since no one is a prophet in his own land, it was at Japan that the Americans W. E. Deming and J. M. Juran, considered the fathers of the concept of quality, aimed their applied ideas (see Weiss, 1988). At first, it was a question of small groups responsible for improving processes and products at the most gênerai levels. But the idea of 'total quality1, literally die quality 'of everything together', soon became indispensable and was treated as a collective goal for which responsibility was shared, rendering the very notion of quality control useless (when speaking of the same thing, American authors tend to use the word 'excellence'). If quality has had such success in Japan, it is also for preventative reasons: Japanese products are exported Worldwide without die need to develop costly networks of after-sales service. Other pillars of the Japanese management System hâve been advanced to explain the success of its companies: first and foremost, employment for life, which in reality oniy involves a fifth of wage-earners but testifies above ail to the attachment and quasi-feudal loyaity of employées to their company. The Japanese does not exist as an individual but as a member of a group, a clan. This is the context in which we shouid relocate another piliar of the Japanese management System: consensus. Group pressure is essential to explain the apparent social harmony which reigns in Japanese society and companies. Certain sayings like 'hammering down 4. I hâve writtcn a book on this subject, with Picrrc-Noel Giraud (19S7)J cntitkd Radioscopie du Japon. 23

For anoiher future

the nail which sticks out' iiluminate the conditions under which the famous consensus is achieved. In addition to the officiai control of hierarchy - 'being beaten on the head' - there is the unofficial but omniprésent controt of colleagues: 'being pulled by tlie feet'. How could people not stay in line under such conditions? In each case it is a question of doing as the others do: not leaving the office before one's colleagues, not behaving differently. Thus the famed consensus is less the resuit of negouation than the fruit of a collective self-censuring where each renounces his or her own ideas for fear of compromising the harmony of the group, which, in any case, would not forgive the m for it. Nor is it at ail surprising that trade unions avoid unreasonable social conflicts and that the 'good' union leaders are rewarded with promotion. Indeed, this occurs to such an extent that, in order to become a manager, it is useful to hâve had union responsibilities at one time or another. Another cliché of the Japanese mode!: the performance of manufactiiring industry thanks to a longer working day than elsewhere. The officiai statistics are more theoretical than real, for the constraints of production and group life are such that, 'spontaneously', many employées stay late in the evening and corne back on Saturdays - or even Sundays - to achieve their targets. Group pressure is such that it is better to do as one's colleagues do and give up a significant proportion of one's leave. Motivation and mobilization are thus at the heart of the System. At Honda, for example, the prize goes to whoever makes tlie most proposais for improving productivity. This motivation détermines the significant bonuses (worth severai months' salaries) which are paid twice a year. Thus, contrary to common wisdom, the Japanese miracle résides neither in management advances nor technology, but rather in the organization of production, in the most concrète sensé of the term. It is a question of doing well what one has to do with order, method, tidiness and, of course, the ail-important wcll-motivated staff. None of this is outside our grasp; the Japanese eiectronic industries set up in the United Kingdom are already performing better than those in Japan itself. However, the United States example demonstrates mat simply transferring a model (or rather its image) to another reality does not automaticatly work: production by the tightened/stretched flovv justin-time method has shown itself to be more costîy than useful in numerous companies. For its part, Germany testifies to another reality: the Germans are as efficient as the Japanese, yet work iess than tlie French, 24

Spreadi>ig the cultures of prospective and strategy

Ultimately, the principal virtue of the true-false Japanese management modet is to hâve unleashed a véritable mentai révolution in our companies (quality circles, company projects, self-training, the cuit of business). Future historians will perhaps recognize Japanese methods, as vvell as the salutary rôle they hâve played in the West.

4. Spreading the cultures of prospective and strategy At the end of this introductory chapter, we are resolutely optiniistic. Indeed, debate and dispute over prospective and strategy are fertile and fashions help models to progress and ideas to advance. We may end by confirming the évidence. If the recipes presented by American business schools and management bestsellers were as useful as claimed, American businesses would be performing better and United States industry less sick. Companies are not the only ones affected by the plotting of a desired future. Authorities, local communities and associations hâve already got the forward-planning itch. The democratization of prospective is only just beginning. It is safe to bet tliat prospective and stratégie cultures will spread widely through organizations. This démocratie diffusion cannot fail to raise the ievel of debate. American soaps and Japanese curios will be abandoned for books by classic authors and reflection on the strong and durable performance models which are burgeoning around us in Germany, France, Italy and Spain. Thèse models are based on long-term, planetary visions and ambitions, efficient management of the (technical, financial, human) means of production to achieve objectives, and highly intelligent ulilization of tlie human lever, and ail done with respect for identity and in harmony with culture. In a word, the movement being sketched out is indeed that 'of a European school of stratégie management.'f As Fernand Braudel said (1980): 'What unités Europe is its diversity.' The common characteristic of this European school is precisely the variety of organizations and behaviours which are proving effective. Thus the principle of contingency is tlie essential basis of this school, which has aiways existed, but which is only just beginning to recognize itself.

5. This was the title chosen by A F ΠT and AFPLANE for their international colloquium held on 1-2 February 1990. Proceedings are availablc frorn AFCET, 16 Boulevard Percire, Paris 75017, France.

25

Far anotlier future

5. One logic in ten chapters The field covered by îhis manual is above al! that of prospective. Developments in company strategy aim primarily to propose a synthesis of widely scattered material. By relying on this synthesis, it wili be possible to illuminate the complementarities anci synergism between the tvvo approaches. In this sensé, this work retains the logic of Scénarios and Stratégie Management (Godet, 19S7). The first two chapters deal with the fundamenta! principles of anticipation. As the future is not written anywhere and lias still to be built, prospective thinking lias nothing to do with the determinism of fururotogy or forecasting. The critique of quantification and extrapolation of trends is necessary to improve the complementarity between approaches: planning needs forecasts with scénarios. The following chapters iist the methods of the prospective tool-box, illustrated by case-studies. In this way we présent the necessary mathematicaf procédures for identifying the key variables of the future^ analysing the interplay of actors, reducing uncertainty and finally building scénarios and defining stratégie options. The third chapter is devoted to the scénarios method, whose logical séquence (délimitation of the System, rétrospective analysis, actors' strategy and constructing scénarios) has become established after dozens of prospective studies. However, this entirely literary logic is an insufficient weapon for tackling the analysis, compréhension and cxplanation of increasingiy complet Systems. Hence the need to call on the more formaiized tools of Systems analysis defined by Barel (1971) in the following manner: 'Systems analysis, in most cases, consists in bringing out the fact that the goal examined nuist be re-placed in a wider context than ils original System. The term "analysis" serves to underline the usefulness of breaking down complex problems into their constituent parts.' • ïn the chapters which foilow, we shal! présent some of thèse formalized tools to which one can rurn at one stage or another of the scénarios method. • Structural analysis and the MICMAC® method, particularly useful for délimitation of the system and for determining the essential variables, are presented in Chapter 4. • Analysis of the past and of future projects is a valuable support for understanding the interplay of actors whose analysis by the MACTOR® method is the basis of Chapter 5. • Morphological analysis, useful for scanning the field of possibilités 26

One logic in lot chapters

and identifying the key dimensions of the scénarios, is presented at the beginning of Chapter 6. • Expert méthodologies (DeSphi and cross-impact) allow us to assess die likelihood of différent hypothèses acting on the variables and key dimensions for the future. Chapter 6 is speciaily devoted to one of thèse: SMIC methodology. By presenting thèse différent tools and mediodoiogical supports, we will at die same time better understand how to utiiize die scénarios method. Let's be clear, however, diat, although the séquence is logical, it is not essential to follow it from A to Z: everything dépends on die degree of knowledge one has of the System studied and the goals being pursued. The scénarios mediod is a modular approach and can, where necessary, be limited to the study of particular modules, for example: • Structural analysis of, and search for3 key variables. • Analysis ofactors' strategy. • Surveying experts on key hypothèses for the future. One of the biggest constralnts of the scénarios method is time. In gênerai twelve to eighteen months are needed to carry out die logical procédure in its entirety; half of tliis time wiil be spent on construction of die base. If only three to six months are availabte, it is préférable to concentrate one's thoughts on the module which seems most important. The complète scénarios method has only been applied in just under half die cases quoted in Chapter 3. In order to iilustrate die modular character of the tools of prospective, they are presented via différent case-studies. Ail the saine, for reasons of confïdentiality it has not been possible to présent a complète and detailed example in ail stages. The outlook is clear, the scénario System will continue to play a référence rôle but will seldom be^arried through from A to Z. That can only be the business of specialists undertaking long and exacting work. The prevailing winds are blowing in the direction of a democratization of prospective. Prospective is itself moving to embrace the appropriation of methods and is imposing openness and simplicity. This is a strong trend which ought to reinforce still furdier die enthusiasm for prospective workshops and to favour the modular use of tools - as needed, and as appropriate to pariicular probSems, circumstances and tempéraments. It is dien appropriate to identify and evaluate possible stratégie options. Chapter 7 présents the problematic of the decision-making and évaluation process and develops the principal methods of multicriteria choice available, notably the Multipol method, a pardcularly simple and easy-to-use tool. Chapter S concerns itself with marrying prospective and strategy (by showing how the meeting and integraùon of thèse two approaches 27

For another future

was inévitable). This chapter also provides the opportunity to return to the sources of strategy and management. On the way, we discover that thèse concepts are as old as the hills and that modem authors hâve invented nothing - the essentials had already been said and written a long time ago by forerunners such as Henri Fayol, Frederick Winslow Taylor and Mary Parker Follet. One lesson may be drawn froni this rétrospective: there are a thousand and one keys to excellence. Chapter 9 attempts a synthesis. It demonstrates that people are at due heart of the différence between organizations that win and those diat lose. It also describes the metamorphosis of structures and behaviours that is indispensable if old habits are to be broken and a true mental révolution is to be iaunched. The final chapter, entided 'The Dawn of the Twenty-First Century', proposes a review of the main trends and uncertainties which organizations will hâve to face in the future.

28

2. From determinism to détermination, from forecasting to 'stratégie prospective'

'The future cannoi beforecast, it is prepared' Maurice Blondel, philosopher

Prospective is neither forecasting nor futurology, but a mode of thinking for action and against fatalism. Prospective recognizes that the future is the 'raison d'être of the présent', that it is multiple and uncertain, and that prospective is a way for everyone to take controi of their own futures. This chapter looks at the rôle of such qualitative prospective analysis in the context of business planning and forecasting. Scénario-building émerges as more vital than ever for forecasting and stratégie management.

1. Action-orienteci anti-fatalistic thinking Those who claim to foretel! or forecast the future are inevitably liars, for the future is not written anywhere - it is stil! to be constructed. This is fortunate., for without this uncertainty, human activity would lose its degree of freedom and its meaning - the hope of a desired future. If the future were totally foreseeable and certain, the présent would be corne unliveable. Certainty is death. Thus, the first aim of prospective thoughtis to illuminate the choices of the présent by the light of possible futures. Good forecasts are not necessarily those vvhich are realized, but those which lead to action to avoid the dangers and arrive at the desired objective. The fréquent errors that occur in forecasu'ng and the notable absence of forecasts of crises bear witness to the crisis in forecasting itself. The impossibility of seeing ahead into the future solely by means 29

From deternunisnt to détermination, front forecasiing to 'stratégie prospective'

of using past data explains the futility of classical économie models3 which fait to include qualitative and unquantifiable parameters, such as projects entertained by, and the behaviour of, the principal actors. The future must be studied with a view to illuminating the présent; that is the basic idea which inspires la prospective. Such prospective anaiysis involves taking a view which is globalj qualitative and voluntarist (a concept created by G. Berger in the late 1950s). It is neither forecasting, with an excess of deterministic quantification, nor futurology, a concept fashionable mainly in the English-speaking world and which embraces ail aspects of research into the future without spécifie référence to die criteria of globatism and vvill. Futurology has had rather a bad press in Western Europe, where it is widely seen as a throwback to the crystal bail or a variant of science fiction. At the same time, the concept of prospective anaiysis has found acceptance mainly in the Latin world (southern Europe, Latin America). Another significant différence between the two cultural worlds is that the concept of 'technological forecasting' has not been accepted on the European side of the Atlantic as being capable of providing a sufficiently global insight into social developments; rather, it is seen as restricted to trends in science and technology and their conséquences, on the implicit premise that science is no more than a product of society and as such cannot suffice to explain it. The concept of forecasting is not self-contained, but is subject to fluctuation from one period and from one society to another. To oversimplify, there are threc attitudes to the future (passivity, adaptation and voluntarism)j in practice, hovvever, thèse constantly overlap and intermingle. The passive attitude to the future is a iegacy of religious fatalism. The 1739 édition of Richeiet's dictionary contains the following: 'Prévision: se dit de Dieu et signifie connaissance de ce qui adviendra!

According to this view, the future is inévitable and virtually predetermined, having already been written by the hand of God, so that man has no alternative but to submit to his destiny; this being the case, it serves no purpose to encumber the présent with future misfortunesj since tomorrow wil! be tomorrow and there is nothing that can be donc aboutit. With the advent of the Industrial Révolution and the sharp accélération in the pace of change during the twentieth century, people came to realize that the pages of history, whether or not written in advance, were tiirning faster and faster. If change was inévitable, it was perhaps aîso predictable, in which case it coutd be worth while 30

Aciion-orietued anti-fatalistic thinking

anticipating it so as to be ready for it when it came, and to take advantage of the opportunities chat it offered. An adaptative attitude to the future thus developed as a reaction to rapid change, bringing in its train a phénoménal development in économie, technological and social forecasting accompanied by excessive and sometimes blind confidence in the new gods of mathematical model-making and econometric metliods. This adaptative attitude is essentially reactive rather than pro-active, resigned ramer than voluntaristic, since it has not purged the original sin of determinism and inevitability. Curiously, fatalism has cast off its religions habits to assume the mantle of a new idol - science and technology. Technological forecasting seems to hâve become the modem substitute for divinely inspired prophecy. The passive Svait and see' attitude has been rransmuted into blind faith in the unlimited powers of a technological good fairy, who will soive ail problems, including those for which she is responsible. Such technological optimism, which has now become so widespread, has also triggered a reaction generating its antithesis technological pessimism (the world is doomed, \ve are playing with fire3 the problems that technology cannot solve are those which are of its own making), which is just as heavily impregnated with determinism. In the face of the accelerating pace of change, the uncertainties of the future, and the increasing complexity of phenomena and interactions, résignation is no longer an acceptable response. An antifatalistic, pre-active and pro-active attitude is essential. People are not fated to accept relationships which are necessary, determined and outside their control. They create reiationships through their own wili and their actions. At the root of ail prospective is an assumption of freedom in the face of multiple and îndeterminate futures. Prospective has nothing to do with the determinism of futurology and crystal-ball gazing. Nor is prospective the same as forecasting, which is too greatly affected by quantification and extrapolation of trends. Prospective does not see the future simply as a continuation of the past, because the future is open to the games of many players, who are acting today in accordance with their plans for the future. So, prospective is a way of thinking which throws light on présent action by looking at possible futures. In modem society, anticipation is imperative because of the combined effects of two main factors: • Firsdy, the accélération of technical, économie and social change, which nécessitâtes iong-term vision: 'the faster you drive the further your headlights must shine.' 31

Front deterntinism to détermination, from forecasiing la 'stratégie prospective'

• Secondly, factors of inertia inhérent in structures and behaviour mean that \ve must sow the seeds of change today if \ve wish îo harvest them tomorrow. Although the world is changing, however, the direction of this change is uncertain. For a company or local authority, change carries numerous uncertainties (économie, technological and social) which it must intégrale into its strategy. Prospective does not daim to eliminate this uncertainty tlirough illusory prédiction, but aims simply to reduce it as far as possible, and to take décisions based as little as possible on hypothetical futures (which is a lot in itself).

2. A critique of forecasting 2 . 1 . D E FIN' 1 T I O N S

Confusion between projection, forecasting, prospective analysis and planning is a source of many forecasting mishaps and misunderstandings. The following définitions should serve to reduce such problems (see Table 1). • Projection is the extension into the future of past developments using certain assumptions for the extrapolation or variations of trends. A projection constantes a forecast onfy where it is based on probability. • Forecasting is the assessment, vvith a degree of confidence (probability), of a trend over a given period. The assessment will generally be expressed in figures and based on past data and a number of assumptions. • Exploratory prospective analysis is a panorama of possible futures, or scénarios, which are not improbable in the light of past causalities, and the interaction between the intentions of interested parties. Each such scénario (a cohérent séries of assumptions) may be the subject of an assessment expressed in figures, i.e. a forecast. • Planning, to quote Ackoff (1969), 'consists in conceiving a desired future and the practical means of achieving it'. It therefore calls for a normative prospective approach. It should be made clear that the plan (an instrument of discipline and consistency) is only one stage in the planning process (an instrument of dialogue). The plan must be simultaneously informative (diagnosis), indicative (remédies) and directive (ends and means). Too many people fall into the classic trap of confusing planning with forecasting, mistaking a divergence from a predetermined objective for a forecasting error. 32

A critique of forecasting

T A B L lî 1 . Summary définitions of the main concepts used in the literature on the future Conjecture

Prospective anaiysis

Probable hypothesis

All-embracing, qualitative, and quantitative^ voluntarist, multiple scénarios (targely a Latin concept)

Projection

Forecasling

Eixtension into the future or variation of past trends

Assessment with a degree of confidence: quantitative, deterministic

Prédiction

Fit

ut

rology

Statemcnt of fact before the event (Delphic oracie)

Al! aspects of research on the future (an Anglo-Saxon concept)

Prophecy

Scénarios

Prédiction by divine inspiration

Cohérent séries of assumptions Planning

'consists in conceiving a desired future and the practical means of achieving it' (Ackoff, 1969)

The historicai paradox is that forecasting deveioped at a time when it was easiest and ieast necessary. In practice, econometric models, backed up by computers, were being used to demonstrate things that anyone could easily find out for himself; everything was more or less direcdy related to the gross national product; GNP itseif was rising by about 5 per cent a year. Basically, time was the best 'explanatory' variable. The year 1973 marked the turning-point at which the future ceased to resemble the past, making the need for a prospective anaiysis (to take account of breakdowns in trends) more immédiate. Such breakdowns are associated with new behaviour patterns, so that models based on past data and relationships are powerless as predictors. This is not the place for a detailed survey of the causes of errors in forecasting, but attention can be drawn to three factors: • Inaccurate data coupled with unstable models. • Excessive quantification and extrapolation. • Explanation of the future in terms of the past. 33

Froni detenninism to deiennination, from forecasting to 'stratégie prospective"

2 . 2 . INACCURATE DATA AND UNSTAlîLE MODELS

Economies does nor obey rules as rigorous as those applied in the more 'exact' sciences such as physics. Statistical data are often supplied vvithout any estimate of the extent of errors which may hâve been made in compiling them. What purpose is served by sophisticated calculations made to the nearest hundredth or thousandth when the figures to the left of the décimal point cannot be relied upon? In the United States, according to O. Morgenstern (1963), even the GNP figure can only be regarded as accurate to within 10 per cent either way, and there is no reason to suppose that errors are always in the same direction; even the direction of variations in GNP may be in doubt. Economists - and others - should always ask themselves about the sensitivity of their results to minimal changes in input data.

2.3.

E X C E S S I V E Q U A N T I F I C A T I O N AND E X T R A P O L A T I O N

Most forecasting methods are based on extrapolation of trends, which assumes that 'ail things will be equaP in other sphères. This is quite unrealistic when the environment is increasingly subject to change and the phenomena to be taken into account are increasingly complex and interdependent. In thèse circumstances, uncertainty about the future is reflected in divergences from a trend. This makes econometric models powerless to anticipate major longterm structural changes, although their users regufariy attxibute their forecasting errors to 'hidden' variables. Piecemeal forecasting which takes account of only a few (usually économie and quantifled) explanatory variables and ignores changes in balances of power and the émergence of new trends is more misleading than useful. Its înaccuracy is attributable largely to the fact that the économie sector is becoming self-contained, so that économie forecasting is divorced from social and political forecasting, while at the same time it is tending increasingly to be broken dovvn into technological forecasts, population forecasts and so on. It is therefore ciear that it has become necessary to take a global view; the pace of change is accelerating, there is increasing interdependence and interaction, and nothing can any longer be taken for granted. The time has corne for forecasting based on quantification to be replaced by an overall prospective approach taking account of ail the qualitative parameters (whether or not quantifiable) which may hâve an immédiate or remote impact on whatever phenomenon may be under examination. 34

A critique of forecasting

Given its relative importance, the anodyne notion of stabilily requires careful study. Various viewpoints are possible. Lot us suppose a static framework. Given a matrix A where coefficient a(ij) represenîs the numberof goods ; produced in région/ Each a(ij) is purchased in this région in the proportion x(j). We therefore make the unrealistic assumption that for x(j) is an average value j, independent of / so that the national output of goods / effectivety consumed is bfi) = X a(ij) X (j). The following problems can be raised: given that A = (a(ij)), b = (b(ij)), what wi!l be thc value of A? If A varies by (A)/i, what wilî be the corresponding variation (A)À' of A? Or again, if b varies by (S)è, what will bc the corresponding variation (S)Arof A? In other words, the problem is one of the siudy of the stability of A'as a function of the minute variations (à)AoîAaxu\ (5)6 of b. We need not recall the rather restrictive theory of 'appropriating' matrices for problems involving linear équations. The following wcll-known example amply illustrâtes this:

A=

7

S

7

32

5 6 5

6 10 9

s

23 33

7

7 5 6 5

8 fi 10 9

0

7

10 7

S 7

10 •7

A=

A + AA =

S

b=

9 10

7

32

5

23 33

9 10

6.99 4.99

+6b=

9

9 9,98

1 1

31

S. 10 7.20 5

7.S 5.45 6 S 5.98 9.S9

X—

1 1

X+ SX=

9.2 -12.6 4.5

3!

I.

32 23

-81 137 -34

33 3!

33 .

\. Whatever the problem under considération, it is essential to carry out a fui! study of the stability factor without which it is impossible ta give significance to the results. 2. Mostj if not ail, statïc linear models are in fact devoid of interest; because of the way real-word values are subdivided, the coefficients obtained are known with a margin of error so large as to make the final results usetess. Source: Bruter, 1977, p. 58.

35

From deterniinistn to détermination, Jrom forecmtitig la 'stratégie prospective'

Adopt a global vision for local action

Prospective thinking must be global - hardly any problems can be isolated. In fact, vve are witnessing quite the opposite trend - a grovving interdependence., even entanglement, of problems. We cannot hope to find solutions which are not global, even if the point of application of thèse solutions is local. Moreover, the complexity of éléments and reiationships to be caken into account, and the need to put them into a global perspective, makes analysis particularly difficuit. This is why it is necessary to use methods of Systems analysis.

2.4. T H E FUTURE EXPLAINED IN TERMS OF THE PAST

The prospective approach accepts tliat there îs a multiplicity of possible futures at any given time and that the actual future will be the outeome of the interplay between the various protagonists in a given situation and their respective intentions. How the fumre evolves is explained as much by hurnan action as by the influence of causalities. The future should not be seen as a single predetermined iine, an extension of the past; on the contrary, it is plural and indeterminate. The pluraiity of tlie future and tlie scope for freedom of human action are mutually explanatory; the future has not been written, but remains to be created (see Fig. 2). Future

Présent Past

Possible futures

Past

Présent FÏG.

36

2.

A critiqua o/forecasn'ng

The prospective approach has evolved from the realization that the future is the product of both causality and freedom; that which is undergone in the future is a resuit of past actions (and of course unforeseeable natura! phenomena); that which is willed explains présent actions. In other words, the future is explained not oniy by the past but also by the image of the future imprinted upon the présent. For instance, an individual's consumption at a given time is not determined solely by bis past income (savings), but also by the future income that he anticipâtes (crédit), as Milton Friedman has ciearly demonstrated in his theory of continuing income. The eye of the future illuminâtes the présent; seen in this light, the future is today's raison d'être. TABLE 2

Classical forecasting

Prospective approach

Viewpoint

Pièce mcal 'Everything elsc being equaP

Overall approach 'Nothing else being cquaP

Variables

Quantitative, objective and known

Qualitative, not necessarily quantitative, subjective, known or hidden

Relationships

Static, fixed structures

Dynamic, evolving structures

Exportation

The past expOins the future

The future is the raison d'être of the présent

Future

Single and certain

Multiple and uncertain

Method

Deterministic and quantitative models (econometric, mathematical)

intentiona! analysis; qualitative (structural analysis) and stochastic (cross-impacts) models

Attitude to the future

Passive or adaptive (future cornes about)

Active and créative (future brought about)

37

Proin deierministu lo détermination, from forccasting la 'stratégie prospective'

3. Usefulness and scope of prospective 3.1.

USEFULNESS OF

AND

PROSPECTIVE

CREDIBILITY ANALYSIS

The usefulness of prospective analysis dépends iargeiy on the aims pursued by tliose who conduct or commission such studies. The objectives that prospective analysis can serve include: Stratégie objectives • Guidance of présent action (the plan) in die light of possible and désirable futures, for example, by aiming for maximum flexibility in the face of uncertainty. • Staking out possible futures so as to be able to take a sélective view of the multitude of current events, identifying those which point to the future, i.e. 'events whose présent dimensions are minute, bue whose potential conséquences are enormous' (Massé, 1965). Tactical objectives • Using warnings in an attempt to pre-empt or foster future developments (inflation rate, économie growth rate, etc.). • Testing a hypothesis or theory, for example, in order to justify a décision and forestail possible criticism, • Culrivating a purposeful and dynamic image. • Developing communication within an organization or between it and the outside world. • Chalienging received ideas and unconstructive behaviour (scepticism, blind failli, shortsightedness). Given the wide range of objectives pursued by those who carry out or commission prospective analysis, little purpose can be served by inquiring into its credibility; prospective analysis is not neutral, but dépends on subjective choices regarding the approach employed with regard to spécifie problems, the assumptions tested, the aims pursued. Because of the uncertainly of the future it is necessary to make judgements and then back them; only after a séries of such judgements has been proved right or wrong can the credibility of prospective analysis be assessed.

38

Usefuhiess and scope of prospective

3 . 2 . WHAT INGREDIENTS GO INTO MAKING AN ACCU RATE FORECAST? An immédiate observation is that the 'correct' forecast is not necessarily the one vvhich turns out to be true. When we anticipate a future problem, we normaily set in motion actions that ensure that it will not take place or that its effects will be mitigated. A second observation is that an accurate forecast is not necessarily a useful forecast. In forecasting, as in statistical testing, three types of error can be identifïed: • Retaining a hypothesis which ultimately is not substantiated. • Excluding a hypothesis or an event that does in fact transpire. • Not asking the right questions, i.e. not incorporating in the prospective chinking the hypothèses or events mat are going to play a prédominant rôle in the future. This third risk occurs reguiarly in forecasting and can best be expiained by the 'lamp-post effect' - where the drunk who has lost his keys looks for them under me iamp-post, since that is where the ground is lit.

3.3. W H A T MAKES A GOOD E X P E R T ?

We hâve a simple answer to this question, but none the less it does not solve the problem: the best experts are generally a minorityj this leaves the question open as to who, among the minorities, has it right. In practice, in many fïelds, particularly those of a technological nature, expert prédictions are often the main source of information. The success of methods like Delphi (where a convergence of opinion is sought through a somewhat directed, even manipulated consensus) can readily be expiained. Such methods are appropriate when it comes to obtaining décisions through consensus, but they are more often misleading than useful as forecasts.1 As an illustration of this point, the reader should refiect on the results of an évaluation of experts made several years ago (see box).

\. One should bcar in rnind that 'convergence' is not necessarily {i.e. rarely) indicative of smtistical consistenevj as we shatl sec when we devetop the SMIC Prob-Expert® cross-impact method. 39

Front detenninisni to détermination, front forecasting la 'stratégie prospective'

An assessment of experts' forecasting performance In a 1976 Futures article, George Wise, of General Electric's R&D centre, analysée! the accuracy of forecasts made in the United States between 1890 and 1940 on technological progress and the impact of innovation. His analysis was based on 1,556 forecasts of technological advances (some of which materialized, while others did not) in eighteen spécifie areas of technology: and the anticipated social, économie or political effects of thèse changes. He drew five conclusions: • Experts produeing long-term forecasts (ten years and upwards) arc wrong more often than they are right. • Experts are slightly more aecurate in their forecasts than non-experts. • Forecasts made by experts in a field outside their speciality are no less aecurate than those of experts specializing in the field concerned. • Forecasts concerned with the continuance of the technological status quo at a given moment arc no more aecurate than forecasts concerned with technological innovation. • The effects of technoîogical development are less casy to forecast than the developments themselves,

Some guidelincs for prospective analysis • Do it in-house rather than rely on outside bodies. • Select a 'new consultant' in préférence to one of the traditional type. • The traditional type seeks to shock the client with his preliminary diagnosis, making him feel guilty so that he allows himself to employ an outside team organizing a major assignment. • The new type seeks to become a catalyst for communication and new ideas within the undertaking, setting up and helping to run in-house prospective analysis workshops or studies broadening the horizon beyond the confines of the undertaking. • Approach problems frorn several angles, according to their complexky. • Bc sceptical of certainties.

3.4.

T H E ABSENCE OF NEUTRALITY OF I N F O R M A T I O N AND F O R E C A S T S

Information about the future, the présent or the past is rarely neutrai: it usually serves spécifie interests. How many studies and analyses are lying hidden in drawers because they are politically undesirable? How many pertinent reports are emptied of tlieir substance by the sélective use of words? How many reaiities are ignored because they do not fit preconceived ideas? Information is manipuiated simply because it is a major source of power. 4Q

Pluralism and complementarity

To inform someone, to give him information which he does not possess, is to divest it from oneself, to relinquish thosc trump cards which could be soid and to make oneself vulnérable in the face of takeovcr attempts, Total communication is impossible. To enter into a relationship with another person, to seek to open up to him, is ai the samc time to hide and to protect oneself behinci fortifications and to oppose him. Briefiy, any relation with another person is stratégie and contains a power component, however buried or subtimated it may be (Crozier and Friedberg, 1977). Prospective information does not escape from this influence and the conséquent pollution; there seems to be an unending stream of alternately alarmist and reassuring reports on population, environment3 energy, arms, etc., ail mutuatly contradictory in their conclusions. By its nature, prospective information is relatively more durable, but is less of a burning issue than headline news. Nevertheless, \ve scan thèse futures to illuminate and guide présent action. The future is an emerging landscape with unknown contours; the constraint is that, despite the unknown horizons, we hâve to take décisions today that commit us for the future. Even if the information is degraded we hâve to take out our bets now, to create the future rather than to submit to it.

4. Pluralism and complementarity 4 . 1 .

PROBLEM SAMD

M ETHO D S

The problem of spécification, identification and calculation of the parameters for a model are classic problems for econometrics. Three problems in particular bave received insufficient considération, despite the fact that the manner in which they are solved conditions die validity of any forecast. The problem ofidentijying expkmaiory variables Let the variable to be explained be x, whose devetopment is to be investigated and forecast. The First immédiate question is: What are the variables that hâve to be taken into account in order to explain x? Some of thèse explanatory variables will be readily identifiable on the basis of thorough knowledge of their past influence on the variable, while others will be more difficult, either because the nature of their influence is not sufficientiy well known or because it bas not yet manifested itself fully. 41

From deicrnntn'sm to détermination, front forecasting to 'stratégie prospective'

It is not sufficient to identify thèse variables. It is necessary to specify the form of the relarionship function (linear, exponentiaL power, iag, etc.). This requires drawing upon theoreticai knowledge, expérience and intuition. It calls for a good knowledge of the past (analysis of corrélations) but, above ail., imagination and creativity. This is why the problems of specifying and identifying variables and relationships also necessitate the use of qualitative methods (e.g. brain-storming, synectics, data analysis, morphological analysis, structural analysis). The problem of estimating future values of explanatory vaiiables

When the explanatory variables hâve been specified and identified, and the parameters of the function hâve been estimated, the next problem in the forecasting process is to make a quantitative estîmate of the future values of the explanatory variables. This problem can be expressed as follows: either, what will be the value of each variable at instant r, or, alternatively, by what date will each variable hâve reached a given magnitude? This type of question is often dealt witli by extrapolating or more or less arbitrariiy fîxing the values of the explanatory variables, whereas in theory each of thèse variables should be explained together and in turn. However, this would make for an endless and heavily looped model and; given the constraints imposed by known data and availabîe time and resources, it would not be possible to go very far along such lines. This is why estimâtes of the future values of explanatory variables are generaliy obtained by the use of simple or sophisticated methods of extrapolation (analogies, logistical curves); in short, it is as though each explanatory variable explained itself over time, with time acting as a pseudo-explanatory variable. The problem of uncertain estimated values

An estimate does not become a forecast until it can be supported with a certain degree of confidence. In coping with this problem of establishing the probability, the uncertainty, the forecaster has little to go on apart from the intuitive judgements of experts when applying methods of the Delphi or cross-impact type. Sélection of forecasting methods

No miracle forecasting metiiod capable of soiving ail problems exists. In practice, because there are so many différent types of forecasting problem, each forecaster will hâve a more or less complète box of tools 42

Pluralism and complcmentarity

from which he makes a sélection according to the nature of the problem ar hand, his objective and the constraints (data and time available) within which he has to work. In the absence of an all-purpose tool, a 'kit' has to be made up from the tools available to meet the requirements of each forecasting problem, an approprtate methodological response which shouid include improvisation wherever necessary. Finally, it shouid be remembered tliat, in industrial prospective analysis, the choice of a method for forecasting the development or sales of a product dépends on certain criteria, such as the life cycle of the products concerned (birth, growth, maturity, décline - see Figure 3). It is found, for instance, that conventions! econometric method s are particularly appropriate forecasting tools for products in their mature phase, when plenty of past séries are available. In the early stages of a product's development, on the other hand, methods of the Delphi or cross-impact type are more appropriate (note: more appropriate, not 'appropriate').

43

Front detcrminism to détermination, from forecasting to

'stratégie prospective'

4 . 2 . P L A N N I N G N E K D S F O R E C A S T S \V l T M S C E N A R I O S

The dangers of excessive quantification (the ever-present tendency to concentrate on things which are quantifiable to the détriment of those which are not) should not lead to a rejection of s ca ris tics, but merely to cautious use of them. The statistics (mathematical, econometric) contained in forecasting models are essential to an assessment of the conséquences of scénarios. The validity of a forecasting model is determined by that of the assumptions (économie, political, etc.) upon which it is founded; the purpose of prospective analysis is to fill in the background, providing the sets of assumptions which will give the mode! its validity, that is, its consistency with future reality. The complementarity between prospective analysis and forecasting has led to the création of 'new forecasting'. The contribution of prospective analysis meets three essential needs of forecasting; • The need for explanation; détermination of known or hidden essentiai variables makes for improved sélection of indicators. • The need for assumptions: construction of scénarios - cohérent set of probable assumptions based on explanatory variables - is the process which lends validity to a forecasting model. • The need for quantification: forecasting according to scénarios provides a means of quantifyîng the results and conséquences of prospective analysis and taking account of that which cannot be quantifkd. Using the following notation of the above model y =f(X-)s the contribution of prospective analysis is to supply, in the form of scénarios, consistent sets of probable assumptions based on the explanatory variables K- and the function / refattng them. Thèse assumptions provide the background for application of a forecasting model to enable calculation of the desired final resuit, y. This 'new forecasting' is a synthesis of prospective analysis and forecasting, which are two complementary approaches; it incorporâtes the rules and constraints of econometric techniques in the cohérent framework provided by scénarios. This scénario is outiined in Figure 4. Researchers would therefore do well to make the distinction between forecasting and prospective analysis more often. The value of a forecast is no more than that of the underlying assumptions. More often than not, severa! cohérent sets of assumptions (scénarios) can be considered fairly probable, so that a forecast should never be published without giving an indication of the estimated probabiliiy of the corresponding scénario. Attention should also be drawn to the dangers of only considering the scénario that is îhought to be tlie most probable. Finally, it should be noted that some scénarios may correspond to breakdowns 44

Plunilism and complcmentcirily

in past trends. Such scénarios are ouîside ihe scope of projection and cannot be calculated according to the conventional forecasting mediods.2

Fi G. 4. New forecasting.

4 . 3 . PROPOSALS FOR S 0 LVIN G THE CRIS! S IN FORECASTING

There can be no sensible planning vvithout exhaustive considération of possible futures and debate on désirable futures and the means of bringing them about. If this debate is to be organized on a proper footing in the light of ail relevant information, statistical forecasts corresponding to each of the alternative scénarios envisaged must be availabie on a regular basis. It is at the interface between prospective considération of possible and désirable futures and quantification by means of forecasting models that the greatest difficulties are encountered in national and corporate planning exercises. Hovv can statisdes be produced in die face of uncertainty? How can scénarios and models be integrated with each other? Thèse are two of the questions which inevitabiy arise; as yet there are no satisfactory answers, and they are eminendy suiîable subjects for research.

Ifonc takes N assumptions, there will be 2A'sets of possible assumptions covering a period lcading up to a given date, the sum of the probabi!ities of the corresponding scénarios being equal to 1. Assume that N = 2 and that the four possible .scénarios hâve the fbllowing probabilités: SI, probability 0.4; S2, 0.25; S3, 0.1 S; and S4, 0.20. Clearly, the most probable scénario (SI) is in tact the leust probable outeome - the most probable outeome ss that any one of the three other scénarios will be borne out. 4S

Front determiuism to détermination, from forecasting to 'stratégie prospective'

As a contribution to the necessary research effort, the following comments are offered to the researchers concerned. First, economists are too inclined to forget that a model is not realiiy, but one of a varieiy of possible means of gaining an insighi inw realiiy.

They should therefore make it their practice, in the interests of prudence and neutrality, to broaden their methodological options (by testing several models and différent approaches., each acting as a failsafe System for the others). It is not acceptable to restrict the range of possible futures in the name of logic and constraints deriving from a model built on the past. Variants should not be determined according to the nature of models, but scénarios should be taken as the starting-point for the construction of models which are most appropriate for the représentation of the development ïo be explored. In other words, it is truly essential to adjust the models to the reality to be constructed rather than the reverse: the means should not be allowed to outweigh the ends. Second, the fact that a model is sophisticated and contains thousands of stmultaneous équations does not necessarily mean that it gives a better description of reality. Such exhaustiveness is ail the more misleading when factors which are unquantifiable, not well known, or controversial hâve been left ou t. In fact, such complex models are becoming increasingly incompréhensible to users who hâve not been involved in their development (trade union représentatives, company economists, researchers working in other fields, etc.). This detracts further from the réputation of modelling, which has already suffered seriously as a resuit of the large numbers of forecasting failures. This defect in forecasting is felt particularly keenly by public and private undertakings which hâve to take décisions affecting their future by feeling their way without any clear frame of référence or cohérent objectives. Thèse undertakings are disarmed because they hâve nowhere to tu m otlier than to forecasting centres which are ill-prepared or illplaced to quantify the uncertainty resulting from the muitiplicity of possible scénarios. Third, some peopie take the view that, since the standard planning period is five years, long-term prospective analysis (ten years or more) is a lower priority than short- and medium-term forecasting. This opinion is groundiess, since the timescale of prospective analysis should be determined by that of the plans of interested parties or of possible breakdowns in trends, which could occur in the near future. Présent action should be guided by the possible futures, since, when fog and uncertainty abound, an excellent compass-setting is a fundamental prerequisite for survivai. 46

Practical advice for future thinking

4.4. KEY CONCEPTS FOR A

METHODOLOGY

One can conclude that quantitative forecasting and qualitative prospective studies are two approaches where tlie 'qualitative' and the 'quantitative' must be complementary. Constructing a forecasting approach by scénarios such as those proposed hère is not a simple task, since there is a lack of methods. This methodologicai gap refiects insufficient past research into tlie topic as well as a breakdown between the two networks - that of tlie forecast model-builders and that of the 'prospectivists', i.e. the futures analysts. In practice, it would be désirable to bring together economists, sociologists, historians, etc., on the same forecast, even though the models may match less well with figures than with reality. In short, the global view of prospective must be based on pluraîism and complementarity of approaches. This is the price to be paid if forecasting is to find its way out of crisis.

5. Practical advice for future thinking How to implement future thinking? Wc can recommend from expérience four ways to improve the relevance and efflciency of prospective analysis.

5 . 1 . A S K T il E R I G II 'I" Q U E S T I O N S A N D M ï S T R U S T C O N V E N TI O N A 3., W I S D O M

Everybody remembers Woody Allen's famous remark: 'The answer is yes, but what is the question?' As there is no right answer to a wrong question it is essential to pose real questions. Unfortunately common wisdom générâtes numerous false problems, i.e. clichés proved wrong by the facts. Light créâtes sliadow. If present-day projectors are so powerfuily trained on certain problems, this is in order to hide more effectively otlicr problems which people prefer not to see. Today's dominant conventional wisdom and fashionable ideas must be regarded with mistrust, for they are generally a source of error in analysis and forecasting. In order to see things more clearly and to ask the right questions., one should not hesitate to think against the flow, even if this means displeasing others. This is an important challenge, for there is little chance of prescribing tlie appropriace remedy if one is mistaken over tlie diagnosis. 47

From detcrminism k> détermination, from forecasting to 'stratégie prospective'

In gênerai, che majority of experts are conformists (it is easy to rake refuge in the majority and leave the task of explaining to others) and conservadve.-1 Very often, good forecasts, that is, those which see things accuratety, are achieved by a minority of experts who demonstrate daring and imagination. The most difficult thing is, of course, to know how to pick out tlie 'good' minority point of view from the rest. For the practitioner of prospective, the lesson is clear: although it may be difficult and risky a priori to défend one new idea among other possibie ones, he must not be afraid to attack dominant ideas. It is in this spirit that \ve were able, for example, to: • announce as early as I97S both a new oil crisis and the energy surplus which would render nuclear energy less necessary^ • dispute the thesis of the industriai décline of Europe in relation to the United States and Japan;5 • cast doubt on current clichés about the Japanese model;s • question the often-evoked thesis of Gcrmany's décline — at a time when Germany has become the world's leading exporter of manufactured products;7 • denounce the myth of technology as the lever with which to émerge from crists.N The day before yesterday (in the 1970s) energy was blamed as tlie cause of ali our ills; yesterday (the 1980s) technology was going to save us; now, the priority is training (renamed, for today's circumstances, 'nonmaterial investment'). After having fallen victim to the technologica! mirage, we are now suffering from 'diploma diseascV It seems that human societies pass from one mirage to another in order to avoid confronting the forest of real problems, which are of a socioorganizational nature: technological and educational changes impty social change and a révolution of the mind.

3. As O. Clarke emphasb.cd: 'Whcn an eminenr and aged expert says that something is impossible, there is a good chance that lie is mistaken.' 4. Energy Horizon Kceps Reccding?', The Times, 5 September 1978, 5. Le déciin industriel de l'Europe', Le Monde, 21 Deccniber 19S2; 'Le vent en poupe', I^e Monde, 5 Fehruary 19S6; 'The USA: Recovery or Concealed Décline?', Futures, Vol. 17, No. 3, 1985, pp. 196-201. 6. 'l'en Unfïishioniible and Controversïal Findings on Japan', Futures, Vol. 19, No. 4, 1987, pp. 371-84. 7. Sec also: 'Germitny: Faasdoxical Power', Futures, Vol. 12, No. 4, 19S9, pp. 344-60. S. From the Technological Mirage to the Social Iîreakthrough', Futures, Vol. 17, No. 3, Î9S6, pp. 369-75. 9. Worldwide Challenges and Crisis in Education Systems', Futures, Vol. 20, No. 3, 19SS, pp. 241-51. 48

Pracîica l advice for finit rc thinking

This point applies equaliy to companies. It is often in structures, behaviour and quaiity of personne! that we should seek to understand the situation of companies in difficulty. Otherwise, how are we to explain the fact that other companies, facing the same market and environmentai constraints., enjoy a high degree of success? Over the last two décades we hâve also noticed that errors of forecasting are often based upon two mistakes: • Overestimation of the pace of change (of technologies). • Underestimation of inertial factors (structures, behaviours). Therefore, when thinking about the future, we suggest it is useful to start by identifying factors which are unlikely to change.

5 . 2 . T H I N K [ N T It Ii L O M G '•[• E R M A N D I M A G E N E F 1RS T W M A T W I L L N O T C H A N G E

Fernand Braudel repeatedly demonstrated the necessity of long-term vision (encompassing several décades, or even severai centuries) in understanding the evolving rhythm of societies, économies or ecoSystems. For example, fourteenth-century Europe experienced a séries of épidémies and famines which eaused the total population to fall by about 40 per cent, but which also halted the excessive deforestation eaused by the previous rapid population growth. This major crisis was to be followed by the Renaissance and finalty, several centuries later, by the progress of the Age of Science and Enlightenment. In a sensé we are still being carried along on thèse waves of innovation, which hâve continued to accumulate for almost six centuries. So crises - the conséquence of rigidities - are often favourable moments for undertaking necessary adaptations and wide-reaching reforms which would otherwise hâve been held in check by multiple résistances to change. Crises bring opportunity for social and organizational change; unfortunately, in many cases crises are necessary for change to happen. History does not repeat itself. With the passage of time, however, societies retain disconcerting similarities in behaviour, which lead them, when faced with comparable situations, to react in almost identicai, and consequently foreseeable, ways. Most events which are destined to happen bave already taken root in the distant past. So the past holds forgotten lessons, which can teach us much about the future. A good prospective study therefore almost always implies considérable anaiysis and rétrospective thought as a preliminary. This may represent two-thirds of the time invested in such a study. A priori, 49

Frnw determiiiisiu to détermination, from forecasting to 'stratégie prospective'

the past has the advantage of being less cluttered and calmer than the présent; however, it too is liable to many différent interprétations. Thus, looking at the past remains a difficuit task, for it means reconstituting scattered and often controversial information. We should not underestimate the importance of factors of inertia in relation to forces of poientiaî change. In prospective, there is often the tendency to imagine what could eventually change, while forgetting systematically to record what has at least a good chance of remaining unchanged. Large organizations are characterized by strong inertia, which means thatunless there are discontinuities, changes inevitably take a long time. Tliis makes it imperative to prépare for changes weil in advance.

5 . 3 . U S E SI M P L E , ADAPTAI? L G M E T H O D S T o S TI MU LA TE TMOUGHT AND FACILITA TE COMMUNICATION

There is no universal tool; no one method is a panacea; available data are both overabundant and incomplète. Furthermore, a model is not reaiity, but a means of iooking at reaiity. Ail thèse considérations lead us to point out that the scope of each method or model is relative. The imperfection of the tools, the inaccuracy of data^ and the subjectivity of interprétations are unavoidable realities, which prompt us to opt for piuralîstic and complementary approaches. As far as possible, the results of a model should be tested for their sensitivity to a variety of data inputs and to the use of another tool. Onîy sufnciently robust results should be considered crédible. The main interest of mefhods is not only that they provide results, but also diat they should be the occasion for structured thought and intelligible communication on a given thème, From this point of view, our recommendation to researchers and practitioners is ciear; the container matters little so long as one is intoxicated by the content communication. The most important thing in a study (vvhether it be prospective, strategy or marketing) is not so much the resuiting report as what has happened in the minds of those vvho hâve been invoived in the thought process it has engendered. This fact has important conséquences for practising researchers and advisers, whose work should increasingly consist of 'making peopie do, radier than doing', of letting each person discover the problems and solutions in his own language radier than revealing ready-made trutlis in a scholarly but alien language. This new advice is to former advice 50

Practical advice for future thinking

what psychoanalysis is to conventional medicine. It relies on appropriate methods; i.e. methods which are simple enough to remain adaptable. What is too simple is srupid and wrong, what is too complex is useless. The idea is to use tools which are simple enough to be appropriable by the users and customers of die results. Such appropriation is necessary to turn anticipation into action (see Fig. 1). Stratégie management must use relevant, consistent and likeiy scénarios in order to better adapt and reach the stratégie goals. To construct scénarios and stratégies, we need appropriate tools. For this reason, we hâve elaborated a toolbox which classifi.es problem-solving metliods as follovvs: • Asking the right questions and identifying die key variables: futures workshops and structural analysis with MICMAC® method (see Chapter 4). • Analysing txends and actors' stratégies: rétrospective srudies and MACTOR® method (see Chapter 5). • Reducing uncertaindes to likeiy scénarios: morphological analysis3 expert mediods (Delphi, cross-impacts) (see Chapter 6). • Identification and assessment of stratégie options: multicriteria anaiysis and Muldpol method. Last but not leasi, we hâve good news for users: most of thèse tools are now available on diskettes (PC and MAC). However, users must be cautious and choose proper tools for each problem. Researchers too often apply the same tool indiscriminately to any problem just because they know it!

5 . 4 . S T A R T W IT I-i V R O S P E C T I V E W O R K S M O ? S

In most cases, the collective prospective approach begins widi a one- to two-day residential seminar which proves tlie best vvay to address the problem under study. The working group is introduced to tools and mediods that may prove useful and the group members can become acquainted with each other. They not oniy receive training in metiiodology, but diink productively about die problem under saidy. After die seminar, die group members are in a better position accurately to assess the problem and to adopt an appropriate working mediod. This method vvill not be fully validated undl several weeks hâve elapsed. The rules of the game are easy: the group divides into subgroups, each choosing as its thème one of die foilowing four topics: • Andcipating and acdng on change. • Eiiminadng erroneous preconceived ideas. 51

From dctcriniiiiitu to détermination, from forecasting to 'stratégie prospective'

• Arbitrating between short- and long-term policy décisions. • Compétence rrees: past, présent and future dynamics. The Framework for the vvorkshops can be restricted to just a few overhead transparencies. The workshops can schedule their work sessions into two- to four-hour modules. An exampie of a prospective workshop in action: the Vierzon tovmship facing its future

Af'ter thirty ycars of local communist governmcnt, ail indicators were in the red. The inhabiîants were deserting the town, as were the nrms, with record leveSs of unemployment and strike action; urban planning was haphazard. In June 1990, local town-council élections brought in a new majority (an intelligent bSend of dissident socialists, determined libérais and grass-root ecoiogists). A fortnight later, the new town council decided to give the inhabitants a say and to organise extra-municipal commissions on the town's future. For rwo days some eighty persons (town council officers, trade union représentatives, industriel executives, shop-owners, teachers, senior citizens, civil servantSj etc.) gave diought to the future of their town and drafted some proposais. The two plenary sessions attracted about 300 people and the committec delegates presented their ideas and suggestions, both in the public session but also to the média and to the town couneii members: a mobilisation process was under way, with the aim of seeking a better future for Vierzon. This process can and should continue. It would be a pity merely to be content with the inteliectual stimulus arisirsg from the workshops. From mis point on the image of Vierzon is no ionger the same. Previously it was simply witnessing its own décline; now it intends to contrat its future. More to corne . . .

Prospective workshops may involve groups ranging in size from ten to 100 persons who hâve 'common life expériences' and who wish to think together about possible and/or désirable changes, with a view to controliing them better and re-directing them. T h e idéal number of participants would appear to lie somewhere between twenty-five and thirty-five. It is then possible to hâve a iimited number of subgroups, each with five or six persons, sharing the work assignments. It is advisable to hâve aï least two subgroups working on 'anticipation of change' (so that they can compare their findings) and there must always be one group working on 'preconceived ideas'. In this way certain unspoken assumptions will be voiced and this particuiar group wiil play the rôle of watchdog (and punchball) for ail the other groups. Workshops can serve as 'launching-pads' for new thought processes and as a way to take control of change. In certain cases, however, the usefulness of such workshops can be even more immédiate. Thus, for a 52

Practical advice for future thinking

consuiting group, it may prove advisable to translate future challenges (as identified by the prospective workshops) into associated objectives, to identify subobjectives, possible actions, projects and other studies that would contribute to thèse objectives (using the 'tree of relevance' method), and to compare this potential overview with the reality of the studies and projects under way. This approach also enables rapid identification of any ongoing actions whose aims are imprécise, as well as major challenges which require nevv action. Stratégie prospective workshops

(organisations involved with the author since 1985) • Fédération for

• Renault (automobile consortium) • Bongrain (agrofood) • Chanel (cosmetics) • L'Oréal • EDF-GDF (electrical and gas utilities) • Frcnch Army Command • Ministry for Education, France

Private Education, France • Moselle Régional Amenities Board • Montpellier town council • Vierzon town council • Sollac (steel products) • Adidas . Bull and ICL

The summing-up sessions which follow the workshops enable everyone to gain a better insight into the key problems that should be given priority treatment. This enables the group to define a commonly agreed working metliod (i.e. whether they should use given tools or not) which is compatible with time constraints, available resources, leadtimes and finally with the agreed objectives.

5.5.

PRACTICAL GUIDELINES FOR P R O S P E C T I V E C R O U P S

Composition and modus operandi of the group

The group mandate must be carefully defined (outlining the problems set, the aims, die source of the study, the target audience and the leadtimes). Draftingof intérim documents and the final report should be the collective responsibility of the group. The group should ideally be steered by a 'pilot', chosen from among the group. The pilot's rôle wilf be to lead and co-ordinate the assignment of individual tasks among the group members. The pilot wili be required 53

From detcrminism to détermination, from forecasting to 'stratégie prospective'

to provide leadership throughout the entire study, i.e. he should not be changed. It is his rôle continuously to revievv lead-times and to make necessary adjustments, primarily revising means and tools rather than objectives. Meetings: scheduling and contera

The best frequency appears to be: three or four closely spaced meetings to begin vvith, then at least three annual meetings once the study is under way and finally three to four meetings to conclude. It is also advisable to establish in advance the schedule of meetings and the agendas by about five to six sessions. Every meeting must hâve an agenda and minutes rnust be taken. For each session, ail group members should file a status report for each session on the work for which they are responsible. At the end of each meeting, individual assignments should be detailed and the agenda for the following session approved. Compétence, group action and 'subconîracting'

One should consult specialists before seiecting méthodologies, and particularly before launching an application; similarly, one shouîd consult external and internai experts (for technical expertise and operational specialities); each interview shoutd be conducted by two team members and the conclusions written up immediately. If need be, certain spécifie technical (or other) points could be subcontracted out. Rôle of the group leader and occasional external experts

It is not ahvays easy to put together a study group (bringing together a wide variety of talents and tempéraments). The most difficult point is choosing a group leader capable of investing a considérable amount of his availablc time, both during working hours and off-duty, should this be deemed necessary. The group leader prépares the meetings, takes notes, drafts the minutes, and must involve ail members of the group by assigning the individual tasks. The group leader must also plan ahead and draw up a programme for the ensuing phases, identifying the problems and the most appropriate methods as he or she sees it. It is in this way that the external expert's rôle can prove useful. His input is not only methodological; he can be calied upon to react to group ideas and even to be provocative on occasion. The présence of the 54

Practical advice for future thinking

external expert is not required for ail meetings and it is the group leader who détermines when it is most relevant to call upon his services. Choice ofmcthod: efficiency, motivation and communication

There is no a priori choice of method; however, making a choice is vital if the meetings are to be effective. Where there is no method, there will be no common language, no valid exchanges, no cohérence, no organization of ideas. Method is not an end in itself and one should not be shackîed by its results; method is merely an aid to make the group's thinking more relevant. A rigorous method is also conducive to group cohésion and motivation; intérim status reports - vvhich should be circulated - and visibility of progress also contribute towards motivation. Lastly, the choice of a method will be dépendent on the type of problem under study (cf. start-up seminars), on time constraints and on the simple wish to see results published. The tooSs should be sufficiently simple to remain within the grasp of both the end-users and the target population. Efficient running of meetings is not something that can be improvtsed. To quote Dr Victor Bataillard - author of the COM-Tables - tlie foilowing common déviations vvere noted: • Endless discussions on minor points which make participants lose sight of the main objectives. • Those who direct the debates prefer listening to themselves rather than listening to others. They are convinced that tlie meeting was a success, whereas others in the group hâve been unable to express their views. • The debate focuses on points which are unciear and which interest only a fraction of the audience. • Major contributions are not included in the minutes, nor recorded in any other way, especially when they are not initiated by the most important person présent. • The people running the meeting do not keep to tlie agenda. • The people in charge of the meeting prefer indulging in sma!l-talk rather dian running the meeting. In order to run a meeting efficiently, several methods can be used, such as COM-Tables, METAPLAN and CREAPLAN. They ail hâve points in common, apart from their Germanie origin (a concern for efficiency, even in créativité') • Foi' example, the utilization of coloured cards of varying shapes and sizes (rectangular, ova!, circular), of green cards for positive ideas and orange cards for critical points, of adhesive stickers, etc. .. .

Frotn deierminism to détermination, from forecasting to 'stratégie prospective'

According to such methods ail participants are allowed to express themseives freely but briefly in writing (no more than three lines or seven words per card) or in a présentation not exceeding 30 seconds. Ideas that are beyond the scope of the thèmes under debate are posted up for later discussion. Différences of opinion are identified by a lightning symbol. Each group member contributes to the logistics for the entire group. The results - as posted up - to a large extent reflet:t the ensuing minutes of the meeting.

56

3. The scénarios method

The scénarios method which we discuss hère was largely developed ac the time the author was in charge of the Department of Futures Studies with tlie SEMA Metra Consulting Group, from 1974 to 1979. With hindsight, it is now clear that setting up such an approach in prospective, with twenty or thirty actual applications undertaken in companies and public administration, contributed to: • stimulating stratégie thought and communication within companies; • improving internai fîexibility of response to environmental uncertainty, and providing better préparation for possible System breakdowns; • reorienting policy options according to the future context on which their conséquences would impinge. In the course of this chapter we revievv the origins of scénarios methods, define the main concepts on which they are based and set forth the aims and the logic of die underlying process.

1. Origins and définitions The future is multiple and several potential futures are possible; the path ieading to this or that future is not necessarily unique. The description of a potential future and of the progression towards it comprises a 'scénario'. The word 'scénario' was introduced into futurology by Herman Kahn in his book Lan 2000 (The Year 2000) pubiished in 1968, but the usage there was primarily iiterary, imagination being used 57

The scénarios meihod

to produce rose-tinted or apocalyptic prédictions previously attempted by audiors such as Anatole France {Island of thc Penguins) or George Orwell {1984). In France, die OTAM team vvas die first to use a scénarios method, in a study of geographical futures undertaken for DATAR (1971). Since that time, this method has been adapted to industrial futures, notably in a study of 'chemical-agriculturaP futures carried out by C. Kintz and G. Ribeil at SEMA in 1977.' The US researchers Gordon, Helmer, Dalkey and otliers hâve deveioped several radier more formai methods to construct scénarios. Ail thèse are based on discussions among experts: Defphi, cross-impact matrix, etc. Developments are regularly published in the journals Futures2 and Technological Forecasting and Social Change*

In practice, diere is no one scénarios rnethod, but radier a variety of methods of construction (some simplistic, odiers sophisticated). However, a kind of consensus seems to hâve been reached; the term 'scénarios method' only applies to an approach which includes a number of spécifie steps (Systems analysis, rétrospective, actors' stratégies, élaboration of scénarios) which interrelate as discussed belovv. The foilowing concepts can be defined: » An invariant: a phenomenon assumed to be permanent up to die horizon studied. Example: climatic phenomena. • Sirong trend: a movement affecting a phenomenon in such a vvay diat its development in time can be predicted. Exampie: urbanization. • A germ: a factor of change hardiy perceptible at présent, but which wili constitute a strong trend in tiie future. In fact, a germ variable is what Pierre Massé (1965) described as a harbinger of the future: 'A sign which is slight in terms of présent dimensions but huge in terms of its virtual conséquences.' • Actors: those who play an important rôle in the System through variables which characterize their plans and which they, to some extent, control. Example: the consuming countries, the producing countries, the multinationals, etc., are actors in the energy System. • A strategy: a set of tactics (set of conditional décisions) determining each actor's acts relative to his plan under every possible contingency.

1. See also 'Three Scénarios ibr Hmploymem by thc Ycar 2000', in 'Employaient or thc Obsession with the Future', Fmuribks (spécial issue, 1977). 2. Published in the United Kingdom by Burcerwonh/Heinemann. 3. Published by Elsevicr, New York.

58

Origins and définirions

A conflict: may resuit from die confrontation between opposing stratégies of the actors, and may take the form of an outbreak of tension between two trends (overcrovvding and lack of space, constrained tirne and free urne). The outcome of thèse conflicts détermines the évolution of the balance of power between actors, or strengthens the weight of one trend or another. An event: the notion of an event is defined by E. Borel in the following manner: 'An event is an abstract entity vvhose only characteristic is to happen or not to happen.' An event can be considered as a variable takingonly one of two values, in gênerai ' î ' if the event happens, and '0' if the event does not happen; such an event will be called an isolated event. Randoinness, subjective probabilities: we refer to the long-established but still relevant work of Professor Ville (1937), A phenomenon is said to be random4 when it can take a certain number of values, to each of which is attached a subjective probability. We 'can consider the calculation of the probability of an isolaîed event as a subjective judgement, insofar as the considered event is classified in a category of events which subjectively hâve the same degree of probability. It is thus the expert who, in passing his judgement, establishes his catégories.' A subjective probability is a gamble5 which is almost always listed if we consider an event which will in fact either occur (probabitity 1) or not (probability 0), but which must be considered as won if, among ail the events to which we hâve attributed À''chances in 100 of occurring, there are actually À'in 100 which occurred at the given time.'1

4. Any event in the past or future of which we possess only partial information (i.c. we are incapable of confirming or otherwiso whecher the event took or will take place) shall bc considered a random event. To attribut*! it probability to this event is équivalent to assigning to ail or part of the information a number denoting the degree of randomness' (Ville, 1937). 5. A judgement of probability must be capable of embodiment in a gamble and the overall success of a certain number of thèse wugers will bc the measure of the relevance of the judgement.' 6. When a person says: '1 estimate the probability of an (isolated) event at 75 per cent,' we follow Ville in understanding this judgement to mean: 'If you were to note ail the events to which I woisld nssign u probability of 75 per cent, and if you were to observe, Ébr a large number of cases, the freqi:ency with which the supposed event did in fact occur, I predict that the frequency would be in the vicinity of 75 per cent.' 59

The scénarios method

2. Types of scénarios and stratégies Scénario: 'A totaiity made up of die description of a future situation and of the séquence of events which facilitâtes évolution from tlie original situation to this future siatation.' To tliis définition, proposed by J. C. Bluet and J. Zemor ( 1970), vve would add that the set of events must demonscrate a certain cohérence. Classically, a distinction is made between tlie following: • Possible scénarios, i.e. everything that can be imaged. • Realizable scénarios, i.e. ail that is possible, taking account of constraints. • Désirable scénarios, i.e. which fall into the possible category, but which are not ail necessarily realizable. According to their nature or their probability, thèse scénarios may be termed 'référence', 'trend-based', 'contrasted' or 'normative'. In principle, a trend-based scénario, whether or not it is probable, corresponds to the extrapolation of trends at ail points where choices are to be made. Often the most probable scénario continues to be termed 'trendbased', afthough it may not correspond to a pure and simple extrapolation of trends, as its name suggests. Certainly, in the récent past, when the world changed iess rapidly than today, die most likely development vvas in fact the continuation of trends. For the future, however, the most probable scénario in many cases seems to entait a severe breakdown of current trends. The extrapolation of trends can lead to a situation which contrasts markedly with the présent, as was shown by a study carried out for DATAR (1971), published in Metra under the ride 'Trend-based Scénario for France', and by La Documentation Française under the title Une image de la France en l'an 2000, scénarios de l'inacceptable. In this

case the trend-based scénario is based on extrapolation of trends, but is not the most probable scénario. Since then, following the repercussions of this study, a certain amount of confusion of language set in. We therefore propose to use die terni 'référence scénario' to refer to the most probable scénario, whether or not it is trend-based. A contrasied scénario is die exploration of a deliberately extrême thème, the a priori détermination of a fuaire situation. Whereas the trend-based scénario corresponds to an exploratory approaeh towards the évolution of a future situation, die historié contrasted scénario corresponds to a normative, imaginative, anticipatory approaeh; a scénario for a future situation is chosen that generally contrasts markedly 60

Types of scénarios and stratégies

with the présent (for example, coastal France, France with a population of 100 million); tlien one examines in reverse the course of events, tliat is, the evolutionary scénario, which could lead to tliis situation. Practice has aîso given rise to another définition of the contrasted scénario, also corresponding to an exploratory attitude, examining the évolution of events to arrive at a situation. In tliis case, the contrasted scénario is defined as a highly unlikely course of events, and it is precisely ils generally highly contrasted nature tliat makes it uniikely. This is die définition which we shall adopt hère. However, tliis does not mean tliat we are abandoning the normative for the exploratory; in our View, this distinction is only of operative interest. In fact, once the évolution and the situation are described, in one direction or the other, the corresponding course of events is both exploratory and normative. It is among the realizable scénarios, which hâve a higher than zéro probability, that we ftnd contrasted (unlikely) scénarios and the field of development where the Àrca rtrfii of nr t most probable scénarios Area of possible are found. As regards désirable scénarios , / scénarios désirable scénarios, thèse are found somewhere witliin the possible zone, and are not al! necessarily realizable (Fig. 5 and Table 3). Area of realizable scénarios

T A B L E 3 . Classification of scénarios according to probability and ovcrali vision ^^•^^

Probabiiiiy

Vision

^"~\^^

Probabiiity of scénario Likcly

Uniikely

Exploratory, From rhe présent to the future

Extrapolation base référence scénario (may or may not bc trend-based)

Extrapotation-based contrasted scénario

Anticipatory, Imaginative, normative (from the future towards the présent)

Anticipatory référence scénario

Anticipatory contrastcd scénario

61

The scénarios mclhod

Sonie areas of application oft/ic scénarios method (total or partial application by the author since 1975) • Air transportation parameters • Long-haul passenger demand • Aircraft construction • Shipbuilding in Portugal • The world petrochemical industry • Off-shore industries • The European automobile industry • The cosmctics industry • The dairy produce sector • Pairs and exhibitions in France • Distribution of industrial products • Distribution of electrical goods • Demand for public transport • Transportation modes in the Paris région • Demand for environmental products • The nuclear power sector • French water networks • Migration in the EEC • Motivation of business executives

• The Trois Vallées ski région, France • Postal services • The insurance sector • The Banque de France « Household consumption of banking products • Vidéotex • Geopolitical developments • The Sahel région • The Paris région * The William Saurin Company (agrofood) • Aluminium • Personal wcapons • Tourism and leisure • Mail order services • Noise • Power distribution networks • Building • The steel industry

There is often confusion between scénarios and stratégies. While scénarios dépend on the type of vision adopeed (exploratory, normative or resxoprojective) and on probability, stratégies dépend on attitudes adopted in the face of possible futures. Apparently, it is the concept 'normative' which gives rise to this confusion. In the case of scénarios the word 'normative' is used in a rétroprojective sensé, vvhereas it naturally refers to the notion of norms and objectives when we are îalking about strategy. In other words, there are no scénario-objectives, but only stratégies. It is important not to confuse the dimensions or key components of the scénarios (démographie, technological, économie, polilicai, social., etc.) with the configurations that each of thèse components can présent. This is where morphological analysis cornes in. A system made up of four components, each having four configurations, will in principle hâve 256 (44) possible states. How can we navigate in this morphological space without being swamped by the sheer number of possible combinations? One answer is provided by the combined utilization of the methods of morphological analysis and 62

Objectives oft/ie scénarios niethod

probabilization of combinations of configurations (interplay of hypothèses) through the SMIC method.

3. Objectives of the scénarios method Increasing uncertainty, grovving interdependence, the quickening pace of change in certain areas (political, technological, industrial, etc.), and the noticeable lack of action in others (démographie, energy, sociocultural) are ail factors which call for a futures approach when considering présent actions. Specificaliy, the following are required: • Alternative scénarios for future development, along with identification of die associated problems and opportunités, given the objectives which bave been selected. • The possible actions required to remedy such problems or take advantage of such opportunities. • The conséquences of possible actions, given the scénario envisaged and the objectives selected, The scénarios method specificaliy tries to conceive al! possible futures and to explore the paths leading to them in order to clarify présent actions and their possible conséquences. The objectives of the scénarios method are: • To detect the priority issues for study (key variables), by identifying retationships between the variables of the spécifie System under study through systemic analysis. • To détermine, especially in relation to key variables, the main actors and their stratégies, and the means at their disposai for bringing their projects to a successful conclusion. • To describe, in the form of scénarios,7 the development of the System under study, by taking into account the most probable evolutionary path of the key variables and by using sets of assumptions about the behaviour of the varions actors. The scénarios method comprises two phases: the construction of the database; and, on this basis, the setting out of scénarios which lead to the génération of forecasts. We can, in fact, use the techniques of classical forecasting vvithin the framework defined by a scénario to convert it into quantitative terms. 7. Scénarios develop: on the onc hand, the situation of the phenomenon being srudied and its environment for a chosen horizon, as well as the best route to it {référence scénario); on the other hand, the extrême situations within which the phenomenon is located (contrastai scénarios, both pessimistic and optimistic). 63

The scénarios method

The complementarity between the scénarios of prospective and die models of forecasdng work opérâtes in one direction but not in reverse: models hâve to be tested in the framework of die interplay of probable, cohérent hypothèses supplied by the scénarios, and not die other way round. In other words, designating die variants arising from simulation witli a single model as scénarios is not prospective. By taking die différent scénarios into account, vve can then evaluate die conséquences of previously decided orientations and, with tlie aid of muldcriteria metiiods, vve can deduce die priority stratégie aedons to be taken in order to exploit the expected changes, dius helping to devise die stratégie plan (see Figure 6). It will be seen later in diis book diat the choice of stratégie options, taking environmental scénarios (external diagnosis: direaîs and opportunités) into account is preceded by die stages of an internai audit (strengths and weaknesses) and compeddve positioning. See Figure 7 for a schéma of die scénarios mediod described below.

3 . 1 . C O N S T R U C T I O N O F T H E BASE

The first stage attempts to construct the 'base', i.e. an 'image' of the présent state of die system, which will serve as a stardng-point for die futures study. This image musi be: • detailed and comprehensive, both quandtadvely and qualitative ly; • broad in scope (économie, technological, political, sociological, etc.); • dynamic, clearly idendfying past trends and harbingers of die future; • explanatory of mechanisms of change and of actors (movers of die system). The base is constructed in three phases: • The délimitation of the system studied and die gênerai environment (polidcal, économie, technological, etc.). • Identification of the key variables. • Rétrospective and actors' stratégies. Délimitation of the sysiem studied

This forms a very important phase, Care should be faken not to exclude a priori from die field of study diose technical, économie and political éléments tiiat are novv widiout influence on die phenomenon under study, but which might, in die longer term, begin to exercise significant influence on the development of die system. Conversely, one should avoid fatling into the trap of carrying out a futures study for society at large, whatever the subject of die actual study. 64

Objectives of the scénarios method

FIG . 6. From scénarios to strategy.

When delimiting the System, one dravvs up as complète a list as possible of the variables that should be taken into considération, whether quantifiable or not, tliereby providing an overall vision of the System under study and its environment. In this way, we develop a reasonably accurate définition of the System. In order to achieve this resuit, a number of methods are used: interviews, seminars with specialists, brainstorming, checklist-building, etc. One thus establishes the list of variables that apparentiy characterize the System, and one then divides them into two groups: • Internai variables, which characterize the phenomenon. • Externat variables, which characterize the gênerai explanatory environment of the phenomenon as studied in its démographie, political, économie, industrial, technologicat and social context. 65

The scénarios methocl

F i G . 7. The scénarios method. 66

Objectives oflhe scénarios method

The search for the principal déterminants of tlie System and their parameters is implemented by an examination of the direct and indirect effects of gênerai environmental variables (external variables) or the variables characterizing the phenomenon under study (internai variables). The technique used in this research - structural analysis - is a commonly used, valuable tool. We will see that structural analysis demonstrates a hierarchy of variables (both driver power and dependence). Highlighting certain variables connrms our initial intuition and raises questions about other variables that would not otherwise be asked. The rypology of variables enables a better understanding of the system's structure. The explanatory analysis is carried out across the groups of key variables as identified by structural analysis: it consists of a rétrospective and current analysis of the actor's situation. The rétrospective avoids overemphasis on the current situation, tlie resuit of which would be to bias the study witli conjunctural factors. The aim is to identify tlie mechanisms and the leading actors which hâve influenced the development of the System in the past, and also to throw light on the invariant factors in die System and the major trends. Analysing the contemporary situation also identifies the seeds of change within the movements of the key variables, as wel! as the stratégies of the actors behind thèse movements. To that end, the analysis takes into considération not only the quantified or quantifiabSe data, but also the qualitative parameters: économie, sociological, poiiticai, ecological, etc. This analysis - at the end of the database stage - results in the identification of tlie 'actors' stratégies'. This Seads to a confrontation between the actors' intentions and to tlie résultant development of the balance of power between them, which détermines the future. This provides a synthesis (in table form) of the analysis of past developments and the current situation.

3.2.

BUILDING

SCENARIOS

Given tlie influential factors, the trends, tlie actors' strategy and the seeds of change that we hâve touched on in the preceding section, we can set in motion the scénarios method8 by having the evolutionary mechanisms inîervene and by confronting the actors' stratégies (ail possible convergences and divergences). 8. A scénario is the set formée! by the description of a future situation (or 'final image') and the consistent route which interconnecta the présent and final situations. 67

The scénarios method

The choicc offinal images

If the developmental possibilités arising from the problem under considération are characterized by n hypothèses, then there are 2" possible final images.'* The SMIC" method altows a hierarchy of the 2" final possible images to be obtained from the probabilities assigned to the hypothèses ranked in order of decreasing probability. A choice is then made of the image corresponding to the most probable scénario, together with the images of the contrasting scénarios. Allowing for the degree of uncertainty which governs the hypothèses, the method relies essentially on the consultation of experts. This consists of: • Questioning the experts as to the probability of the hypothèses occurring or not: the simple probabilities of each hypothesis together with the conditional probabilities, since the probabilities can be interconnected. The experts (twenty or thirty) are consulted by questionnaire. They are chosen from différent sectors, depending on the field to be explored (government, entrepreneurial, international, university, etc.). • Calculating the probabilities assigned by each expert to the various scénarios possible, and ranking them by hierarchy. • Carrying out a sensitivity analysis: while varying slightly the probability of a hypothesis, the variations induced in the other hypothèses are then observed. Thèse procédures then allovv déduction of the influential and dépendent hypothèses. The results provided by the various experts and the resulting hypothèses are then compared, allowing identification of: • On the one hand, the final image of the référence scénario, which is the image most often quoted among those best 'placed' by the experts, and which corresponds to the most probable overall outeome. • On the other hand, the contrasted images selected from among the images most often cited by the experts and having a significant mean probability of occurrence. The corresponding scénarios

9. If» = 2, there arc two hypothèses, H(\) and H(2), and four possible final images: H(ï) and H{2) occur; H(\) occurs, H(2) docs not occur, and vice versa; netcher

H(\)nor H(2) occurs. 10. SMIC (System and Matrix of Cross Impacts), presented in Chaptcr S on expert consensus methods. 68

Objectives of Oie scénarios meihod

T A B L ]• 4 . Dimensions and configurations of the Interfuturs scénario Dimensions Relations berween deveioped countries

Configurations envisaged Collégiale management

Partial fragmentation between pôles

Internai dynamics Consensus iavouring Rapidly changing Conflict between of devclopcd socictios liîgh growth values and social groups and moderato growth moderate- growth Evaluation of relative productivité

Convergence

Relations between Large inercase in North and South North-South and among économie exchange deveioping countrics

Increased North-South divide

Divergence Fragmentation of the South by regionj in liaison with the dcvelopcd countries

A scénario associâtes a spécifie configuration wiih cach dimension. Thus therc arc thirty-six possible combinaiions hère. Imerfuturs focused ils analysis on some of thèse. Source: Lesourne and Malkin, 1979.

describe an environmentai évolution vvhich is typically ver y différent from that of the référence scénario. Often images that are pessimistic or optimistic (from a desired point of view) are selected. Developments and 'pathways'

At this stage the scénarios are still embryonic, since each is iimited to the outeome of hypothèses. The main problem is now to describe the 'pathway' from the présent situation to the final images selected for the référence scénario and the contrasted scénarios. The élaboration of a scénario usually calls for a division of the period under study into successive subperiods with intermediate images. Naturaliy the number of thèse subperiods dépends on the natural cycles présent in the System. To ensure cohérence of the 'pathways' between the différent images (présent situation, intermediate and final images), the basic hypothèses are worked through thoroughly. They resuit either from the conclusions deveioped progressively (using information gathered from the base, particularly the table of actors' stratégies) or by induction from the fundamental hypothèses. Thus the scénarios method consists of coherently describing the developments as and when they arise, between the présent situation and 69

The. scénarios method

the chosen horizon, by bringing to bear the evolutionary mechanisms compatible with the hypothèses while following the trends of the main variables for the phenomena as reveated by structural analysis. The scénario is completed by a detailed description of the final image (and the intermédiare images if necessary). The scénarios méthode as described hère, represents a path, die logic of which (délimitation of the system, rétrospective anaiysis, actors' strategy, vvorking out of scénarios) has been confirmed by numerous futures studies (see Figures 8 and 9). However, this logic proves to be an inadéquate tool when it cornes to undertaking analysis, understanding and explanation of increasingly complex Systems; hence the necessity of using the more formai tools of Systems analysis as defined by Y. Barel (1971) as follows. In most cases, Systems analysis consists of emphasizing the fact that the objective under investigation must be placed in a wider context than the original system. The term 'systems analysis' underlines me fact that it is usefu! to break down complex problems into their elementary components. In the chapters which follow we présent some of thèse formai tools, which may be needed at various stages of the scénarios method. • Structural analysis and the MICMAC® method, particularly useful for delimiting the system and determining the key variables, are set out in Chapter 4. • The analysis of the actors is the main topic of Chapter 5. • Expert methods (Delphi and cross-impact) allow probabilities to be assigned to the hypothèses for the key variables of me future probabilities. Chapter 6 describes one of them: the SMIC method, • The methods allowing décisions to be taken in the présence of multiple criteria and in an uncertain future are essential to move from prospective to strategy. To û\e extent that multicriteria methods are décision aids, useful in both prospective and stratégie management, they receive spécifie attention in Chapter 7. Through presenting thèse différent methodological tools, at the same time we gain better insight into how to put the scénarios metiiod into opération. However, although the path is logical, it is not absolutely necessary to follow it from A to 2: everything dépends on the level of knowledge one has of the System being investigated and of me objectives being pursued, The scénarios method is a modular approach: one can limit the study to this or that moduie, as required, for example: • Structural analysis and the search for key variables. • Analysis of the strategy of the actors. • Expert inquiry on key hypothèses for the future. 70

Objectives of the scénarios meilwd

F i e . 8.

One of the main constraints on the scénarios method is time: it generally takes twelve to eighteen months to follow the whole path through, and aï least half of this is taken up with construcnng the base. If there are only three to six months available for the study, it is préférable to concentrate on the module which seems to be the most important. The full scénarios method has been applied in under half of the cases quoted in the box on page 73. To illustrate the modular nature of the 71

The scénarios melhod

FIG.

9.

tools of prospective; we présent tliem in a number of case-studies. For reasons of confidentialité', it is not possible to présent a complète and detalled case in ail its phases. The logical pathway which must îead to the présentation of a plan of action consisis of constructing a basis for thought, the working out of scénarios and the évaluation of alternative stratégies. Although the method does not aspire to universality of application and is no sort of panacea, it at least has the merit of being the fruit of expérience; over neariy ten years, more than thirty prospective studies hâve been successiuily concluded on similar bases. We must not expect from a prospective method that which it cannot provide, however sophisticated it might be: that is, that it should tell us what the future reaily hoids for us. No one can do that, because the future is composed not only of the interplay of determinate factors but also of individuals' freedom. However, the scénarios method can provide simple help towards thinking constructively about the future. It can help us (and this is a strong point in its favour) to choose the strategy which, in the midst of ail kinds of constraints, will prove the most likely to bring our plans to fruition while retaining a maximum set of possible advantages.

72

Objectiva ofllie scénarios meihod

Exainple ofa scénario: doubling life expectancy

We could live, on averagc, to an âge of 140 for men, and 150 for women. If we suppose that the âge for marriage does not vary much, there would be an average of at lcast four or five générations alivc at the same time in eacli family. What would the retirement âge be? If it were 65, me number (proportion) of old-age pensioners would become intolérable for socieiy. As this would be impossible, the retirement âge would be postponed. Thus strong expansion wouid be required in order to create full employment, and we would work to 100 or more. But to work for 100 ycars implies keeping abreast of progress: gênerai retraining wouid become a lifelong commitment. Since one cannot envisage learning sciences chat would still be useful 100 or 200 ycars hence, a change of objective for primary and secondary éducation would be needed. It would hâve to provide the basic toois and prépare the intelligence to adapt itself to change by exercises in mental gymnastics (mathematics and Latin, for exampte, as well as exercises in creativity). There would be an imbalance between générations and conflicts (short of radical change in éducation). Hxtra accommodation would be required, since the population itself would double, even though the birth rate might remain stable. There would be profound changes in inheritance: we would not inherit goods or power until nearly 100 years of âge. Power would tend to remain in the hands of thc over-lOOs. But that would be unacceptableand the younger générations would claim their autonomy. They would be treated as minors until they were nearly 50. Thanks to contraception - accepted by ail under the absotute necessity of restraining population growth - women would be free of maternity at around 35. They would thers hâve around 115 years of life left. Hecausc their services would be needed to support the aged they would ail take up employment. Their salaries woutd no longer be considered as incidental and they would hâve access to ail positions. A number of female geniuses would be revealed — ail those who had not hitherto been able to break out of the sufling sociological web that had up to then been the binding ruie. Public opinion would accept them. Ail those men and women who had neither passion nor creativity nor adéquate spirit or soûl would find life too long and would commit suicide. The troublemakers would last for a long time, with a concomitant growth in criminality. Moral independence would be sought by other means: psychiatry would be revolutionized, etc. Source: lost in the mists of ïimd

73

The scénarios method

4. Illustrations Selecting illustrations is problematic. On tlie one hand, many interesting examples of corporate scénarios cannot be published for reasôns of confidentiality. On tlie other hand, most scénarios published by public administrations or international organizations hâve had no impact on their strategy.

4.1. SCÉNARIOS AS TOOLS OF STRATEGY AND/OR MANAGEMENT

As regards companies, \ve need to dravv a distinction between those that are carrying out highly confidential prospective studies for Uie sole use of the directors making stratégie décisions (Lafarge, Péchiney, Mercedes or Nestlé come to mind), and those which are rather using prospective as a tool for collective thought and mobilization of minds (such as Renault, RATP or the Ministère de l'Équipement with the Direction Départementale de l'Équipement) and which make communication compuisory. In the first case - prospective as a stratégie tool - the Lafarge group represents one of the best examples of 'cold' prospective thought which has had an impact on strategy: by the mid-1970s this group had anticipated the décline of cément (~1 per cent annualiy between now and the end of the century), and had decided to invest in a leading-edge field (biotechnology, with tlie take-over of tlie Coppée group) vvhose sole characteristic in common with cernent vvas its low manpovver requirements. Companies of this type are always in step with the future, and hâve already anticipated the conséquences of the création of Europe and the opening up of Eastern Europe. In the second case - prospective as a tool for mobilization and management - the process of involving staff, which thèse prospective exercises facilitate, allows the company to be better informed as it faces change. The restructuring carried out at Renault (a one-third réduction in manpower between 1985 and 1989) was probably facilitated by the new awareness which emerged from the collective considération of company change carried out between 1982 and 1984 - Opération IESC (Industrial, Economie and Social Change) - in which several thousand middle management and supervisors participated. In this case3 transparency is mandatory, and just as there was an IESC notebook, so there are prospective and management notebooks at tlie Ministère de l'Équipement. Thèse notebooks act as sound-boxes for tlie circulation and amplification of ideas. 74

Illustrations

Between thèse two extrêmes lie one point in common and several intermediate positions. The common point is the permanent but cyciical naaire of prospective, marked by strong upbeats every four, rive or seven years. The long-term backcioth must be of sufficientiy good quality to !ast for several years. It is a little like driving a car - by switching onto full beam a few limes, we are able to drive even better with dipped headlights. As examples of intermediate cases of prospective, where it is both a stratégie tool and a directors' tool for mobilization, we can cite the big oil groups Shell and Elf. Both hâve used scénarios for over twenty years and would even go so far as to advertise this practice as one of the key éléments of their stratégie management. We shail not dwell hère on the expérience of Shell. We merely note that they use a fairly rudimentary method of scénario construction. Scénarios hâve primarily acted as a way of uniting and encouraging stratégie activity and of creating a shared basis for thought for me directors of a highly decentralized group. The success of the scénario method lias been greatly enhanced by the oil shocks which were already foreseen in 1971-72. Our impression, after several accounts by Shell management, is that they are concerned above ail to stimuiate the imagination of their strategists and to make them think together, for example by constructing scénarios of a technological society where information would universaily replace energy. Yet it is not for this reason that the Shell group has moved out of the energy field somewhat and into telematics. Although the scénarios hâve a high level of internai transparency (a too! for communication between directors) and a good degree of cohérence (intellectual logic), they seem to be much weaker when it cornes to examining die relevance of questions and their probability. At the methodological level, this work was inspired by a Frenchman, Pierre Wack (1985), whose own inspirational sources came from the founders of the École Française de Prospective, vvho are not well known in Europe and even less in the United States. Pierre Wack has a number of spiritual successors such as Peter Schwartz, who in 1984 took over from Wack within Shell's prospective group, and who left for the United States in 1986 to set up the Global Business Network (GBN), a network of prospective practiu'oners serving an international club of companies. Today Peter Schwartz is popularizing scénarios in the Englishspeaking world, and has even managed to win over his well-known friend, Michael Porter, who has incorporated tlie notion of scénarios in his most récent works. In mis way French scénarios will soon be rediscovered and imported into Europe under tlie 'made in the United States' label. 75

The scénarios methotl

Planning al Shell, 1945-80 The chronologica! ïist beiow illustrâtes how Shell movccl from linear, highly quantified forecasts to the acceptancc of uncertainty and the description of possible futures based on scénario analysis. 1945-55 Physicai planning 1955-65 Planning by project and sclcctivity 1965-72 Unified Planning Machinery (UPM) System 1967 Start of the Horizon 2000 study 1969-70 Horizon Year Planning Exercise (with a fifteen-year horizon) 197 î Expcrirnenting with scénarios at the company's registered offices in London ! 972-73 Introduction of planning by scénarios 1975 Introduction of medium-term cyclical scénarios 1976-77 More thorough study of'societal analysis' in planning 1978-79 Deeper analysis of geopolitical and political risk 1979-80 New examination of the very long term + development of planning capacity within the Group. The method described by Pierre Wack is now wetl adapted. Group planners hâve accumulatcd a great deal of expérience in analysing the crucial factors intervening in their fteld of activity, in identifying the relationships which develop between the various actors présent and in describing the process of évolution of thèse balances of power. AH thèse analyses are incorporated into the cohérent diagrams put forward for considération by the decision-makers. The task of constructing global scénarios is carried out as a service by part of the Sheli Group, which acts as a global environmental observatory. It produces global scénarios of dcvelopmcnts in the économie environment, the energy environment, the oil environment and so on. In thèse scénarios there are analyses of, among other things, démographie phenomena, political dcvelopmcnts, changes in values and lifestyles, technological and économie developments, monctary problems, energy demand as a function of thèse factors, energy supply, the spécifie position of oil, the possible évolution of relations between the producer and consumer nations, the formation of crude oil priée structures and the hypothetical évolution of thèse priées, etc. Starting with scénarios of the global environment presented by the Group, the subsidiary eompanies construct their own scénarios, by studying thosc aspects which are more spécifie to their national environment and which hâve an impact on their activities. Conceived as toots to aid stratégie thoughtj the national scénarios must be multiple and yet not too mimerons; they must be the resuit of thorough and rigorous analysis, yet remain simple and easy to use; and they must cover a broad spectrum of possibilités, while remaining in the reaim of the probable. In 1981 ail the oil eompanies, fearing the conséquences of the confiiet which had just broken out between Iraq and Iran, built up reserves of oil. Shell, thanks to the method described above, got rid of its surpluses before the rnarket became overcrowded and priées coltapsed. 76

Hlustraikn

4.2.

S U C C È S S FUI. U T I L I Z A T I O N OP S C E N A R I O S - T H E C A S E O F E L F

Hère we wave die French fiag with die case of" the Elf Group, which carried out three prospective exercises in fifteen years; the first in 1969 looking ahead to the horizon of 1985, the second in 1978-79 with the horizon of 1990, and die third in 1985-87 with the horizon of 1995. We were fortunate to be associated with the last of thèse., and reasons of conndenuality force us to limit our comments to die memodological aspects and to what François Didier was able to say and Paul Alba (1988) to pubtish. First exercise in 1969: Elfup to 1985

A small team, repordng to die Research Director (Bernard Delapaime) and made up of several internai and external consultants, was setup and produced a report. With hindsight it is interesting to note that although die oil shocks were not foreseen, nevertheless the vulnerability of the Group's situation in oil and in Algeria was perceived. Hence the recommendation diat die Group diversify, both geographically (into the North Sea) and sectorally (into chemicals, pharmaceuticals), which was acted upon effectively. In the field of technologie developments in batteries for fuel (electric) vehicles and aruficial proteins were generally overestimated: two or diree years Iater the Group halted its research in this field. The arrivai of telematics and the growing importance of environmental problems were foreseen, with the création of a centre for information and research into possible harmful effects on the environment. Overall, the success rate of this first exercise is largely positive. It recommended that the Group double its size in relation to its competitors between 1969 and 1985. This objective was achieved. Second exercise in 1978-79: Elf up to 1990

This study, under die responsibility of the Director of Strategy, François Didier, engaged the équivalent of eight full-time staff. The main goai was to define stratégies in the face of three identified scénarios of global development (black, grey and rosy). As in the preceding exercise, developments in die energy field were wrongly perceived: it was at the height of the second oil shock, yet die price of oil in constant dollars fell. On die otlier hand, die study was very prudent as regards technological change. This did not prevent it from emphasizing innovation in the technical froid, as well as in the économie and socio-organizational fields, 77

The scénarios mcOiod

and reorganizing its chemical activities while deveioping biotechnologies vvith Sanofi. Also to the crédit of this study is a recommendation which was effectively acted upon: the setting up of Groups for Reftection and Study on Stratégie Thèmes (GRETS). Around twenty of thèse were operating during the 1980s. Third exercise in 1985-87: Elfitp to 1995 Still under the responsibility of François Didier, the study took eighteen months and invoived several dozen people in the group. It was carried out in two phases: an exploratory phase, in order to construct environmental scénarios; and a normative phase, in order to define stratégies in the face of thèse scénarios. During the exploratory phase about ten working groups were formed according to branch of activity and thème (environment, research, geopolitics). During the normative phase, three sample stratégies were studied: 'natural resources', 'commodities' and 'high tech, high growth'. This exercise, like the earlier ones, was undertaken without a particular methodology^ but with a high level of professionalism in activating the groups and respecting deadlines. The collective mobilization was undertaken in an upbeat manner: meetings with the président, branch directors, etc. I had the opportunity to follow this exercise ciosely in my rôle as external consultant to the pilot commitîee - an exciting and, at times, difficult rôle of observer, I was required to remain silent for hours and to react, if possible in a critical manner, to what had been said and put forward by différent members of the group., some of whom were highlevel managers. The two external consultants were also used sensibîy, as they were placed at the disposai of groups which felt the need for Hvely input, but were not imposed upon them. It is prématuré to assess the success record of this exercise. P. Alba (1988) merely notes as a conclusion: 'Once a company is important it cannot make leaps.' Elf Aquitaine has thus confined itself to basing its future development on its three industrial branches (oil, chemicals and health and hygiène) and on thèse alone. The exercise to 1995 led to a mémo from Président Pecqueur on company direction, written on 17 July 1987 and distributed to ail staff. The Elf and Shell groups make good use of scénarios precisely because of the high degree of internai involvement that they hâve been able to develop. This does not prevent errors in forecasting and stratégie diagnosis, and we therefore believe that thèse groups would be well 78

Anticipation and scénarios: myt/is and rcalities

advised to make a little more use of existing prospective tools in order to strengthen the relevance, cohérence and probability of their analyses. As regards transparency, it would hardly be possible to do more for collective memory whiist observing necessary confidentiality. Intelligent use of prospective and of scénarios (as a tool for strategy and management) has aiso long been évident in companies such as the RATP (Paris public transport system) and for some years at EDF (French electricity board).

5. Anticipation and scénarios: myths and realities The very use of the word 'scénario' can prove dangerous for future thinking: there is a risk of being swamped by média successes with littie or no scientific grounding. Let us examine two preliminary questions: • Does using tlie term 'scénario' for any combination of hypothèses (for a given analysis), however attractive this may be, confer a degree of future respectabilité'? • Need one necessarily draw up full and detaiied scénarios when undertaking future thinking? The answer is most assuredly 'No!' on both counts. A scénario is not a future reality but a way of foreseeing the future, thereby throwing light on the présent in terms of ail possible and désirable futures. The test of reality, and a concern for efficiency, should be used to guide prospective thinking in order to gain a better mastery of history. A scénarios approach can only be crédible and useful if it compiles with four prerequisites: retevance, consistence, tikelihood and transparency. In other word s, one must ask die right questions and cîearly formuiate the true hypothèses which are keys to the future. Without mis procédure, tliere is a risk of leaving out some 80 per cent of ail possible futures. With modem probability tools, such as the microcomputer package SMIC-Prob-Expert®, it takes only a few minutes to provide resutts for a group study. Curiously, certain proponents of me prospective approach refuse to submit their own thoughts on an issue to a system which is akin to a tie-detector, or which would at least reveal contradictions in their reasoning. The last prerequisite mentioned above (transparency from A to Z) impties thaï: 'a clear concept can alvvays be stated clearly'. This should apply to any problem, the methods used to solve it, the reasoning behind it, the results and the conclusions in regard to the scénarios envisaged. Far too often, unfortunately, either die simple reading of the scénarios 79

The scénarios method

proves laborious and the reader must invest considérable effort in ascertaining the prerequisite conditions (relevance, consistence) or the Hterary quality is so low that the reader finds it indigestible and sets it aside. Thus, owing to a lack of close and critical review, a number of scénarios remain crédible, i.e. they are given the benefit of the doubt and the reader is left feeling somewhat guilty that he has not read the text through to a iogical end. Without transparency, resulis will be unadaptable and vvill not enable implication of the actors (the public) that we wish to involve through the scénarios. Naturally, the transparency and attracdveness of scénarios do not preclude quality of content; scénarios with catchy ntles, or which are presented in an emotion-ridden, pleasurable or doomsday style - such as l-'uture Shock - can be convincing. Such works are fiction, that is, a literary genre which per se is quite honourable and often makes for superb reading. A famous example is George Orwell's Ninetecn Bighiy-l'bur. However, they rarely contain relevant, cohérent or even lîkely scénarios. By replying negatively to the second question above, we want to make it clcar that anticipation and scénarios are not synonytnous. Too

many studies in prospective hâve become bogged down over lime because a group decided to launch into 'die scénarios method'. Butwhy, we may ask, did they do so? A scénario is not an end in îtself; it only becomes meaningful vvhen its results and implications are embodied in real action. Undertaking a scénarios approach is dme-consuming (tweive to eighteen months is not uncommon) and there must be several persons involved in order to establish a team context and make the process viable. After three years the leaders of the OECD Interfuturs team (1976-79) announced that they had insufficient urne usefully to exploit ail the results! (see Lesourne and Maikin, 1979). We can safely add an extra year for circuladng and publicizing results. In most corporate and administrative organizadons, such teams will be required to report within the year. In extrême cases, policy-makers may launch a prospective stiidy that they wish to see finished in a matter of a few weeks. Thus the prevailing conditions are rarely ideai and it is better to throw a little light (rather than no light at ail) on die impending décisions. Sheer common sensé dictâtes die simple questions that one shoutd raise at the outset: What can be donc in the given time, using the means available? How can it be donc in such a way as to be bodi crédible and useful to die decision-makers? From tliis point of view, it will often be advisable to iimit die scénarios to several key hypothèses, say four, five or six. Beyond such numbers, die sheer magnitude of possible combinations is such that the human mind 80

Anticipation and scénarios; myths and realilies

simply gives up. Such straightforward scénarios are used as backgrounds for stratégie options such as \vhat if... ?' or 'what for . . . ?'. Short-cuts in a scénarios approach make it ail the more crucial to do some pretiminary thinking about the key variables, the trends and the actors' stratégies. One final difficulty that arises when building scénarios and selecting methodology relates to lead-times. Even if one has months, or even years, to finish the assignment, there is a risk inhérent in the start-up phase because team members or even the team leader may change as the study progresses. A futures study rarely survives after the departure of the initiator. In major organizations — given existing staff mobility factors — it is préférable to limit the length of the project to one year and to plan for intérim status reports. It is also advisable to identify a preliminary exploratory phase, during which the éléments atstake are identified, and a reguiatory phase during which the various stratégie policy choices are defined, in terms of items identified in the preceding phase. Hère we end our gênerai présentation of the scénarios rnethod. In the fotlowing chapter we shall return to each of the stages involved and présent the problems that arise, as well as the tools avaiîable to overcome them. This will involve, in turn, identifying the key variables for the structural analysis (Chapter 4), analysing actors' games with the MACTOR® method (Chapter 5), surveying the field of possibilités with morphological analysis (Chapter 6) and reducing uncertainty by the use of expert metliods (Chapter 6).

8t

4. Identifying the key variables: structural analysis

1. Origins and objectives of structural analysis A System consists of a set of interrelated éléments. The System structure, i.e. the network of reiationships between thèse éléments, is essential to an understanding of its évolution, because this structure maintains a certain permanence. The aim of structural analysis is to highlight the structure of the reiationships between the qualitative variables - quantifiable or not - which characterize the System under study (for instance, a company and its stratégie environment). Structural analysis enables one to describe a System by using a matrix which interconnects ail the System components. This method permits analysis of thèse reiationships and identification of the main variables. With regard to the techniques used, as J. Barrand and C. Guigou (1984) noted: 'Structural analysis is based on Leontiev's input-output matrices, on the theory of graphs and the simulation exercises of operational research carried out soon after the last war in the USA, and in particular by the Rand Corporation to fulfil American army requirements.' Some Systems analysis and technological forecasting methods hâve their origins in defence research. This was the case in France for the method of 'relevance trees' developed by the Centre de Prospective et d'Évaluation (CPE) (Centre for Prospective and Evaluation) of the French Ministry of Defence. This was not the case with structural analysis, however, which seems to hâve been introduced into France by Professor Wanty, who worked for the Belgian subsidiary of the METRA International Group and who taught at die University of Paris Dauphine in 1969 and 1970. Since then structural analysis has become more 83

Idenlifyitig tlic key variables: siructural analysis

widespread through, in particular, die work of Professors R. Saint-Paul and P. F Teniére-Buchot (1974), and our own work at SEMA. Structural analysis has two complementary objectives: first, during the initial phase to obtain as thorough a représentation as possible of the system under srudy, in order, second, to reduce systemic compiexity to its main variables. In his thesis on structural analysis and its dcvefopments, J. E Lefebvre (1982) lists several of its applications: • It can help in thinking about a System in order to build a more elaborate model such as Systems dynamics. • It can be used on its own in order to assess stratégie choices. • It can form part of an overall approach such as die scénarios method, • It can help group communications and discussions or group adhérence to a spécifie objective. In practice, structural analysis has been used in two main ways: • In decision-making (research, identification of the variables on which to act to achieve the selected objectives): the POPOLE model of P. F. Tenière-Buchot (1973) is a good example. • In forecasting (research on the key variables which bear on the future dimension). This use was developed in the early 1970s and we were attracted, in particular, by the development of the MICMAC® method, where die importance of a variable is measured less by its direct interreiationships dian by many indirect interreiationships (Duperrin and Godet, 1973). The next section deals mainly with the latter use of structural analysis in forecasting. This comprises several stages: • Listing ail the variables. • Location of interreiationships within the structural analysis matrix. • Search for the key variables by the MICMAC® method.

2. Listing ail the variables In order to develop as exhaustive a list as possible of the variables which define the system formed by the phenomenon under study and its environment, no research path is, a priori, excluded. Ail brainstorming and intuitive methods are useful hère. The process of listing ail die variables should be assisted, preferably, by non-directed interviews with the représentatives of the actors presumed to be invotved in the system under study. The questions should be open, such as: 'In your opinion, what are the factors which will condition die future évolution of such and such a phenomenon?' 84

Ijocaiit»i of reliitionships within the stniciural analysis matrix

To identify thèse variables, several différent politicaL économie, technoiogical and social viewpoints shouid be adopted, in order to build up files and to organize several sessions of collective thinking. Following the accumulation of raw data, the actuat listing of ail variables is based on aggregating and reftning the data so that a homogeneous list is obtained. Also, given the nature of the phenomenon under study, it is often advisabk to reclassify the variables by disdnguishing between internai and externa! variables; the internai variables characterize the subsystem under study whereas the external variables make up its environment. Finally, detailed explanations of thèse variables are essential to identify the varions interrelationships. Indeed, such an explanation can provide the detailed analysis of everything implied in the définition of a variable. Without the création of this common base, the analysis and identification of interrelations could be rendered impossible or meaningless. Thèse files, once compieted, are kept open and they are updated as and when required. They thus form an information database which can systematically sort data. Below, we describe me scale models built, under our supervision; by J. Barrand and C. Guigou (1984) as a case-study of structural analysis. We concentrate on the déterminants of employaient and unemployment; the analysis was reduced to the forty-one variables shown in Table 5.

3. Location of relationships within the structural analysis matrix Within a systemic world-view, a variable can only exist because of its interrelationships; intuitive récognition of the existence of certain interrelationships enabled us to pinpoint some variables when preparing the list. T A B L E 5 . Déterminants of cmpioyment and unemployment Intentai variables

1 2 3 4 5 6 7

Unemployment rate Unemployment rate for women Unemployment rate for young people Number of long-term unemployed Employaient in services Employment in industry Employment in agriculture 85

Ideiitifying the kcy variables: structural analysis

Environmental variables (extertml)

Economie and financial

Internai

International

Techrtological and industrial

Sociodemographic

Institutionat

8 Household income 9 Rate of économie growth 10 Household consumption i l Inflation 12 Savings 13 Financial situation of eompanies 14 Interest rate (savings} 15 Rates on the money market 16 Money supply 17 Budget déficit 18 Constraints on the external balance of payments 19 International context 20 Degree of opcnness of che economv 21 European harmonization 22 Energy price and raw matcrials 23 Investment for rationalization 24 Investment for expansion 25 Operational productivité' per hour 26 Company competitiveness 27 Technological and tndustrial évolution 28 Working population 29 Employment rate among women 30 Geographical concentration of unemployment 31 Immigrant workers 32 Parallel economy, moonlighting 33 Sociaî relationships 34 Uncertainties for the future 35 Evolution in ways of iife 36 Work législation and régulations 37 Work-sharing and income-sharing 38 Distribution of mandatory wage déductions 39 Total value of mandatory wage déductions 40 Training 41 Work costs (average hourly wage)

Structural analysis attempts to interrelate the various variables in a dual input table (see Figure 10).

86

location of relationships within the structural analysis matvix

Internai variables

Extcrnal variables

j Internai variables

I

n

External variables

III

IV

FIG . 1 0 . Structural analysis.

Before establishing whether there is a relationship between the two variables, the prospective group must systemancally answer the foltowing three questions: • Does variable i causaliy affect variable j, or is this relationship the orner way round, i.e. j to /' (Fig. 11 (a))? • Does i hâve an impact on j\ or does some coUinearity exist, i.e. die third variable k lias an impact on i andj ? (Fig. 11 (b)). • Is the relationship between z'andj direct, or does it operate through another listed variable? (Fig. ll(c)).

(c)

F I G . I I . Threc possible cases suggesting thnt ; has a direct impact on/ Many errors can be avoided by using this procédure when fîlling in the matrix. At présent some variables only hâve a weak influence. However, 87

Jdentifying the kcy variables: structural analysis

they could become much more important in a différent context in the future; this is the case, for example, with the 'new technologies for industry and agriculture' variables. Therefore, we should take into account poteniial relationships which, depending on the case, are to be added to the référence relationships, that is, to those certainly in existence. Normally, filling in the matrix is a qualitative exercise: we only state whether a relationship exists or not, although the strength of a particuiar relationship can also be specified. Thus, severai direct relationship strengths can be distinguished: ver y strong (VS), strong (S), average (A), weak (W), very weak (VW) and potential (P). The term 'potential' refers to incipient relationships which still oniy hâve a weak influence but which couid, in a future context, become very infiuential (for example, the conséquences of technical change). Thus, a certain dynarnic can enter into the structural analysis, and the sensitivity of some resuits can be tested against the intensity of the relationships that are being taken into account. (See Figure 12 for matrix example.) One of two methods can be used to fi.ll in the matrix: • In rows, by identifying the influence of each variable on ail others. • In columns, by recording the other variables infiuencing each variable. Theoretically, it is useful to use both methods and to compare the resuits by superimposing the two compteted matrices in order to identify différences and, consequently, the errors made. However, this practice often proves to be a time-consuming luxury the researcher can rarely afford. Indeed, most structural analyses completed thus far interrelate severai dozens of variables; that is, severai thousand questions hâve to be asked, which takes severai days of hard work. Also, structural analysis is a systematic quesù'oning procédure; without the insights offered by the matrix, many of thèse questions would ne ver be asked. Moreover, new variables are sometimes discovered that nobody had thought of when drawingup the original list of variables. Filling in the matrix is a good technique for stimulating dialogue. Indeed, it stimulâtes exchanges of views and discussions to help create a common language within a prospective group. As it is, the structural analysis matrix remains qualitative. With more time available, with a more refined method and, especially, if we were aiming at a différent objective, we could develop a more quantitative matrix: • Giving a positive or négative value w direct relationships, as in the first experiment of J. C. Duperrin and M. Godet (1973). This can be a profitable exercise and reveals positive (ampiifying) and négative (regulating) feedback. To perform such an exercise without making an 88

Location of relaiionships within ihe stniclural anatysis matrix

effort disproportionate to its anticipated usefuiness, the exercise shouid deal only with a small number of variables (ten or fîfîeen). First, it is not always clear whether an impact is positive or négative; second, the désignation of a particular variable could be open to question because turning variables into their opposites (for instance, increase and decrease) does not automatically alter the value of the row and column relationships. For example, there is the case of consumption as a function of income: as income increases, consumption increases; when income decreases, consumption does not necessarily decrease. • Ouantifying the strengths of variables. The coefficients au can, for instance, be considered as elasticities (the percentage variation in variable j following a relative variation in variable 0- In this case, we assume that we can reasonably estimate such elasticities (ranging, a priori, somewhere between less than infinity and infinity). This means that it is impractical or îoo costly to perform this exercise for a matrix with more than ten or fifteen variables. In any case, it is not very useful arbitrarily to attach a quantitative value between 0 and 10 and then perform caicutations which will hâve little meaning because the accurate value of the starting coefficients is not known (Lesca, 1982). Although such refmements may be interesting in principle, they are ill suited to the future use of structural analysis. The search for key variables requires an overall vision of the System under study and excludes restricting the analysis at the start to several variables, since one is actually trying to identify the main variables from among severa! dozen others. Practical recommandations

Sufficiently considered preliminary thought usually leads us to characterize the System under study by a list of about seventy variables (iess structured considération can easiîy lead us to keep about 100 variables), This point is important in terms of the number of questions to ask, which can vary by a factor of two (5,000 or 10,000). We need to allow a minimum of three months to carry out a good structural analysis, incîuding internai and external interviews. Expérience seems to show that a correedy filled in matrix shouid be between 15 and 25 per cent full, depending on the size of the matrix. Higher levels (30 to 35 per cent) are indicative of excessive filling in, with a secondary relationship having wrongly been considered as direct. The influence of external variables on internai variables is much stronger than the other vvay round, which is not surprising. In the same way, the effects of the internai and external variables on themselves are greater, as one would expect. 89

Identifying the key variables: structural analysis

Influence of:

On internai On external variables variables

Internai variables

20%

10%

External variables

15%

25%

Fi G . 12. Example of normal density of direct relationships per section, The higher the rate of entry of direct relationships, the iess relevant will be the considération of indirect relationships by the MICMAC® method. Thus, if the matrix were filled in 100 per cent, the indirect relationships would merely be a similar multiplication of direct relationships, To fitl in a matrix of seventy variables will require about three days' work by a group of five to ten people. During the first half-day, vvhich is often laborious, the group will barely be able to examine more than four or five variables and their impacts on the whole of the System. In doing tliis a common language and a certain cohérence in world-view are created, so that quite naturally the rest of the exercise proceeds much more quickly and easily (a rate of 1,000 questions per half-day is common for the rest of the process). But the group must be open to debate, and should sometimes mark any doubts with question marks, so that diey may corne back to thèse afterwards. Measuring intensity and taking into account potential relationships also represent acceptable compromise solutions to enable collective thought to proceed without causing excessive individual frustration.

4. Search for the key variables with the MICMAC® method Having developed an exhaustive list of the variables to be taken into account, we now hâve to reduce the complexity of the System and to identify which key variables should be studied first. When dealing with an internai subsystem, related to an external environment, there are two kinds of major variables: first, the external variables which are the most infiuential and the most useful in explanation (the system's main déterminants); second, the internai

Searclifor the key variables laiih ihe MICMAC® mclhod

variables which are the most sensitive to this environment. The environmental variables which do not seem to affect the System under study can be discarded. MICMAC® tries to pinpoint the independent and dépendent variables (the key variables) by building a typology of thèse variables in both direct and indirect classifications.

4.1.

DIRECT

AND

INDIRECT

CLASSIFICATION

(MICMAC®) Straightforward examination of the matrix reveals which variables hâve the greatest direct impact,1 but is not enough to reveal the hidden variables which sometimes greatly influence the problem under study. Indeed, in addition to the direct relationships, there are also indirect relationships between variables, through influence chains and réaction loops (feedbacks). A common matrix comprising several dozens of variables can include several million interactions in the form of chains and loops. The human mind cannot conceive and interpret such a network of relationships. The MICMAC® method,; a System of multiplication of matrices applied to the structural matrix, is used to study the diffusion of impacts through réaction paths and loops; thus, a hierarchy can be developed for the variables: • In order of their influences by taking into account the number of paths and loops of length 1, 2,. . . n . .. arising from each variable. • In order of dependence, by taking into account the number of paths and loops of length 1, 2, . . , « , , . accruing to each variable.

The firsi set of information can be obtained by tirât analysing the direct impacts: the su m of the fîrst Une represents the number of times that variable i lias an impact on the System. This number is the indcpcndencc factor of variable i. Similarly, the sum of thej column represents the number of times; was influenccd by other variables and represents the dependence factor of variable j. Thus, for each variable, an independence factor and a dependence factor can be obtained, to classify variables according to thèse two criteria. MICMAC*: Matrice d'Impacts Croisés - Multiplication Appliquée à un Classement (Cross-Impact Matrix - Multiplication Applied to Classification). This meîhod was developed atthe CEA between 1972 and 1974 byj. C. Duperrinand M. Godet (1973). 91

Identifying the key variables: structural analysis

FIG. 13. Dcterminants of employment and unemployment in France during the 1980s. 92

Searchfor the kcy variables tvith the MICMAC® method

4.2. T H E MICMAC® PRINCIPLE: RAISING THE POWER OF THE MATRIX

The MICMAC® principle is a simple one, and is based on the classical properties of Boolean matrices, which are summarized as folîows. Let us examine the foliowing cxampie where the System of variables can be broken down into two subsystems, S[ and 83, which are not indépendant

since they are interconnected through variables a, b, c. As regards the direct effects: • a is strongly dépendent on subsystem S], • c dominâtes subsystem SjWhen anaiysing the direct effects> variable b is neglected, although it can represent an esscntial clément of the system's structure since it is the relational cross-section point between both subsystems S; and 82-

If variable 2'directly influences variable k and if k directly influences variable j, we hâve the foliowing:

In this casej any change affecting variable i can hâve repercussions on variable j. There is an indirect connection between Î and j. Numerous indirect relationships of the

type which exist in the structural analysis matrix cannot be taken into account in a direct classification. When the matrix is squared, secondorder relationships are revealed^ such as:

93

Idenlifying the key variables: structural analysis

Indeed, A2 = A X A = {a2;} where a2^ = Z/; a 1 ^ . û'/e,When a~ydoes not equal 0, there is at least one k where c^^a1^;- 1, i.e. tliere is at least one intermediate variable k where variable i has an impact on kia^ife = 1) and where variabie k has an impact on variablej'CaW = 1). We can say that a second-order patli goes from i toj; if a1^ - N, there are N paths of second-order length going from i toj via N intermediate variables. In particular, when a2j; = N, there are N circuits (or influence Ioops) of second-order length going through variable i. Similarly, by calculating /I 3 , A4, . . . , A"} the number of influence paths (or influence Ioops), of the 3rd3 4th, . . . , wth order, interconnecting the variables, can be found. Each time mis process is repeated, a new hierarchy of variables can be deduced. Their classification is based on the number of indirect actions (influences) they hâve on the other variables. When raised to a certain power (usually the power of 7 or 8)3 tliis hierarchy proves to be stable. This hierarchy is tlie MICMAC® classification. When the linear sum i a"j.- is raised to a power for variable i (a"2y being one élément of die matrix raised to the power of ri), this means that there are a large number of paths of length n rising from variable i} and that variable z'subjects the other System (or subsystem, when dealing with a block) variables to a large number of influences. Hence the MICMAC® classification can classify the variables according to the influence that they hâve (or that they are subjected to) by taking into account die whole network of interrelationships described by the structural analysis matrix. To niake this less abstract, let us take the following example from J. E Lefebvre's thesis (1982): a System is identified by three variables, A, B and C, which interact in the following way:

The structural analysis matrix can thus be written as follows:

Sum of the cléments in each column 94

Search for ihc key variables willi the MICMAC® method

In this first matrix, the diagonal éléments are always set at zéro: this means that the influence one variable lias on itself is not taken into account, whereas in the indirect effects (updated by squaring the matrix) this influence is taken into account (thèse effects always occur through another variable).

The digit 1 in the first row, first column, means that there is a circuit of length 2 going from A to A. Indeed:

The digit 1 in the second row, first column, means that there is a path of length 2 going from B to A. Indeed: B

It is clear that ail éléments of a matrix raised to the power of 3 show that the paths and loops of length 3 go from one variable to the other. It should be noted, and this has already been underlined, that the row and column classification becomes stable when the éléments are raised to a certain power. But the classifications of this matrix raised to a certain power emphasize clearly the importance of certain variables by the indirect effects (feedback) that they hâve.

Hère, the row and column classifications become stable from the power of 4. 95

Identifying the key variables: structural analysis

We obtain the same stability, after several répétitions, with a starting matrix filled in using 1, 2 and 3, depending on the intensity of relationships. Taking immédiate account of the relationships in this way is understandable in so far as we can consider a relationship between two variables of intensity-level '2' to be the équivalent of two direct relationships of intensity levé! T between thèse variables.

4 . 3 . COM PARIS ON BETWEEN DIRECT, INDIRECT AND POTENTIAL CLASSIFICATIONS

Our concern is to pinpoint the most influential and the most dépendent variables. It is understood that the influential variables are those whose évolution will hâve the gréa test effect on the system, while the dépendent variables are those that are most sensitive to the évolution of the system. In addition to a simple examination of the matrix, which allows us to discern which variables hâve the highest number of direct connections with the system, it is a good idea to pick out the 'hidden' variables, i,e, those which, taking account of indirect relationships and feedback loops, a!so appear to be very important. Variables are thus classified according to the number and intensity of relationships in which they are involved, either in an influential or a dépendent capacity. There are three classifications, direct, indirect and potential, according to the nature of the relationships taken into considération. Comparison of direct, indirect and potential classifications is ail the more interesting in that we can associate an approximate time horizon with each of thèse différent classifications: • Direct classification results from the short- to medium-term interplay of relationships; its horizon often corresponds to less than a décade. • Indirect classification intégrâtes chain reactions which necessarily take longer and the time horizon recèdes ta. the médium- to longterm (10-15 years). • Potential classification goes further than the indirect classification, as it intégrâtes relationships which will not émerge until much later and which will only hâve repercussions on the system in the very long term, Obviously the comparison between the classifications (direct and MICMAC®) can confirm the importance of certain variables and also reveal other variables which were previously thought to be unimportant but which play a leading rôle because of indirect actions. It would be wrong to neglect them during the explanatory analysis. The example 96

Scarch for the key variables with the MICAIAC® metliad

given hère explains how a hidden variable linked to the development of nuciear power in France was reveated (see box, p. 98). Every structural analysis using the MICMAC® method revealed two gênerai points: • (At least) four-fifths of the results confirmed initial expectations, and for many variables the indirect classification does not differ from the direct classification. Therefore, the highest-ranked variables can be selected without too much hésitation, whereas those which seem in ail cases to be secondary can be rejecced. • Between 10 and 20 per centof the results seem counter-intuitive since, in the différent hiérarchies, certain variables move quite noticeably up or down me order. This analysis stimulâtes discussion within the prospective group and new questions arise. Some of thèse question preconceived ideas; while other questions arise because of the ideas themselves. We can still remember the shocked reaction of die management of the French Post Office in 1978 to the classification of the positions of thèse variables with respect to their dependency vis-à-vis the gênerai environment. The direct classification seemed the logical one, me mail traffic variables being tlie most dépendent, whereas the workforce policy variables (manpower levels, wages) or the quality of service were considered as internai controi variables. The MICMAC® classification rurned this hierarchy upside-down. After long discussions, mis a priori counter-intuitive resuit became self-evident: on the one hand, mail traffic since 1973 had been recording quasi-autonomous development, almost independent from variations in économie growth, to which it had been closely correlated in the 1960s, On the other hand, one had to admit that the workforce policy of a public service has practicaily nothing to do with what its gênerai management wants, but is closely linked to overall government policies for tlie public sector, the latter also being connected to the political situation. The policy of upgrading the lower salaries and of creating more jobs in the civil service after May 1981 had strong repercussions on the French postal services: over 50 per cent of the workforce belonged to the lowest catégories, C and D, of the public sector.

97

identifying the key variables: structural analysis

Uncovering hidden variables

The following example is based on a prospective survey of French nuclear power. This survey was carried out in 1972 by the CEA {French atomic cnergy commission). By adopting several viewpoints (politicat, économie, technological, etc.), the think-tank for this survey prepared a list of fifty-one variables which should be taken into account. The results obtained are as follows: Direct classification Position 1 2 S 10

Indirect classification: MICMAC® Sensitivity to external effects

Scnsitivity to externa effccts Rcvolutionary îcchnological invention or dcvciopment

Site problems

15 Accidentai nuclear catastrophe 26

Accidentai nuclear catastrophe

29 32

Revolutionary technological invention or deveiopment

Site problems

51

The variable 'sensitivity to external effects' moved up from fifth to first position. Thus, since 1972 structural analysis has enabled us to foresee how important collective psychology and public opinion reactions would be for the deveiopment of nuclear energy. This évolution is even more striking in the case of the variable location problems for the siting of nuclear plants', which moved up from thirty-second position in the first classification to tenth in the second. T h u s , the kind of problems thaï EDF (French central electricity generating board)had toface ai Plogott hadbeen identifiaialmost lenyears bcfore they became a reality.

98

Searcli [or ihe kcy variables with tlm MICA1AC® method

4.4.

THE

INFLUENCE-DEPENDENCE

CHART

Each variable is associated with an influential indicator and a direct dependence indicator across the whole System. The sum of variables can then be positioned on an infiuence-dependence chart (Fig. 14).

FIG.

14.

This chart can be divided into five sectors: Sector 1: Influentia! variables. Thèse high influential, low dépendent variables are the explanatory variables which condition the rest of the system. Sector 2: Relay variables. Thèse variables, which are both highly influentiai and highly dépendent, are unstable by nature. Any action on thèse variables will hâve repercussions on other variables; this will feed back to the relay variables, amplirying or defusing the initial impulse, Sccior 3: Résultant variables. Low influentiat and high dépendent variables. Thèse resulis variables are infiuenced by the déterminants (sector 1) and the relay variables (sector 2). Sector 4: Low influential, low dépendent variables (near the zéro point). Thèse variables constitute marked trends or factors which are relatively unconnected to the system, having only a few relationships with it. Because of their relatively autonomous development, they are not déterminants of the future of the system and we can therefore exclude them from the analysis without too many scruples. Sector 5: Averagely influential and/or dépendent variables. We can say nothing a priori about thèse 'middle clusters' of variables. 99

Idcinifying ihc key variables: structural atialysis

. . . regarding ihe system's stability or instabiliiy

A iow number of relay variables confers relative stability on the System. In facr, in an unstabie System (cluster of points around the main diagonal — see Figure 15), each variable is both influential and dépendent, and any action on one variable has repercussions on ail the others and in uirn on tlie original variable. The advantage of a stable system is thaï it introduces a dichotomy between the influential variables, on which one can or cannot act, and the results variables which dépend on them. Influential scale

Influential scale

Dépendent scale 1. Relatively stable System Fi

G

Dépendent scaie Unstabie System

. 15.

A major benefit of constructing the influence-dependence chart is that it can check whether what is being explained is related to what is assumed a priori to be an explanatory factor and which should normally be considérée) to be influential. This chart often provides surprising resutts (see Fig. 16). The relay variables of the north-east section are, a priori, the key variables at stake in the System - those over which the actors will fight, because of their unstabie character. Freqnently, among thèse key variables, vve find taboo subjects that nobody speaks about, precisely because they are important. On the other hand, many topics which figure among the declared priorities of the company or organization are often situated in the zone of excluded variables, because they are neither influential nor dépendent. In addition to the scénario exercise, a structural analysis should be carried out as a preliminary to the construction of any econometric forecasting model. This analysis could be usefully completed by the method which breaks down the various parameters into carefully related components so as to reveal related (hierarchical) or unrelated subsystems, and to display die key rôle played by certain relay variables, as suggested by J. F. Lefebvre (1982). 100

Search for the key variables viith the MICMAC® method

Influentiai scale

Inftuential scale + Unemployment rate + Social relationships

+ Unemployment rate + Social relationships

MICMAC® classification

Direct classification Dependcncc

Dependence

Kcy: I: International variables M: Monetary variables R: Tcchnical and économie relay variables FlG.

G: Global économie variables B: Business variables 5 117: Social and work variables 16.

There is no single, officiai, scientific interprétation of the MICMAC® results. The think-tank has to ask the right questions and propose explanations. Thus, for instance, in our example we might hâve wondered why the unemployment rate - the parameter which is influential and has a high level of dependence - seems likely to be abie to explain unemployment through the biasing of social relationships. It might mean tliat unemployment is not onîy a résultant (the conséquence of rigidities) but also a change factor which, when raised to a certain level, affects the social climate and imposes politically motivated measures of flexibility to create jobs and job-sharing schemes. Ail thèse questions hâve to be answered. This is the aim of the next step in the scénario method. This step essentially deals vvith the key variables identified by structural anaiysis, locates which actors are involved with thèse variables and for whom the past, présent and future roies hâve to be carefully studied. Prior to initiating this anaiysis on the rôle of actors, we offer hère a (provisional) conclusion on the value and limitations of structural anaiysis.

5. Value and limitations of structural anaiysis The analytical method described above attempts to identify the key variables (whether hidden or not), to ask the right questions or to analyse counter-intuitive aspects of the system's behaviour. The method should j but not replace, the decision-maker. It does not try to describe 101

Identifying the key variables: structural analysis

accuraîely how the System works, but rather its aim is to pinpoint the main éléments of the system's organization. The results should be used while bearing in mind the limitations of the analysis. • The first limitation dérives from the subjective character of the Iist of variables. The précautions taken offer some guarantee of objectivity, but for practical reasons the number of variables should not exceed a few dozen. Therefore, some subvariables related to one of the problem's dimensions hâve to be regrouped, more or less arbitrarily. For this method, tliis is a disadvantage as well as an advantage because any model arbitrarily favouring the quantitative éléments to the détriment of the qualitative éléments is rejected. • The choice of a typology of relationships conditions the results. Usually, two types of relationship are taken into account: (a) the so-called référence relationship (known reiationships which are true today and should remain so in the future); and (b) the potential relationships (possible relationships which hâve not yet appeared, but whose development is incipient). On the plus side, this typology is simple. On the minus side it considers as potential not only what is incipient., but also what is doubtful or disputed. A more extensive study, having more time and means at its disposai, would classify the relationships and their nature (conditional, causal, technicai, instituûonal, psycho-sociological, etc.). • Processing the structural analysis matrix only détermines the existence or non-existence of various relationships. However, this matrix includes relationships with widely varying intensity. Its characteristics should be borne in mind when interpreting the results. That is why the intensity of thèse relationships should be determined, even it only on a qualitative view, in order to carry out sensitivity analyses. However, we should not walk into die trap of excessive!}' quantifying the relationship in order to obtain a semblance of accuracy. While keeping in mind the above limitations of structural anaiysis, tlie results obtained and their essential value should be recalled. This method, first of ail, is a tool which should be used to classify ideas and to tackle a problem systematically. Since several thousands of questions hâve to be asked, certain questions hâve to be asked to uncover variables which would never hâve been thought of otherwise. First, the structural analysis matrix acts as a matrix of discovery and helps to create a common language within a think-iank on the prospective exercise. Second, since a certain number of feedback effects involving each variable hâve to be taken into account, a hierarchy based on tlie influence 102

Value and limitations of structural analysis

and the dependence o(" thc variables should be prepared for such variables. Thus, the main influential of the phenomenon under study can be revealed. The control variables and the résultant variables thus uncovered help beiter to understand the organization and the structure of the System under study. One can scoff at the fact that 80 per cent of the results obtained confirm initial intuition and are obvious, It should first be noted that obvious points are easy to prove with hindsight, but it is more difficult to forecast accurately what, in the quagmire of preconceived ideas, is undeniably certain and obvious. Furthermore, 80 per cent of the obvions resuhs show that Uns approach is logical and commomensical. This requirement is essenrial to give some crédit to liie 20 per cent of 'counterinniiiivc' results.

103

5. Understanding the actors' stratégies: the MACTOR® method

This stage is essential. Proper prospective thinking cannot be carried out unless there is an in-depth rétrospective study. Notably, this means considering ail the key variables and questions identified earlier, and building up a database (both quantitative and qualitative) which should be as extensive as possible. Ail sources of statistical information should be drawn upon to identify the major evolutionary trends, to analyse past discontinuities, the conditions under which thèse came about and the rôle played by the main actors of this évolution. As in the case of structural analysis, the above information should be supplemented by qualitative interviews with the actors themselves; this approach enables one to identify the main events which point the way to the future, and to gain a better overview of the interplay of events and a better compréhension of the relationships between the actors. It is only when a solid database is available and there is a thorough knowledge of future challenges that the MACTOR® method can be usefully implemented. The future is never totally predetermined - however influential past trends may be, the future remains open to several possible scénarios. The actors in the System under examination possess various degrees of freedom which tliey will be able to exercise, through stratégie action, in order to arrive at the goals they hâve set themselves., and thus successfully to carry out their project. From this, it follows that analysis of thèse actors' moves, confronting their pians, examining the balance of power between them (in terms of constraints and means of action) is essential in order to throw light on the stratégie issues and the key questions for the future (which are the outeomes and conséquences of foreseeable battles). 105

Utidersianding the actors' simiegies: the MACTOR* method

If \ve focus our attention on energy, for example, thèse key questions will be concerned mainly with the priée of oil, the demand for energy, the maintenance or coilapse of solidarify among OPEC member countries, and so on. To take another example, in a 1976 futures study of cosmetics consumption up to 1990, anatysis of actors' stratégies showed that the existence of companies with specialized distribution was threatened by the moves of other actors (such as mass distributors., consumer movements and trendsetters). In me field of prospective there is général consensus on tvvo points regarding analysis of actors' moves. On the one hand, everyone concurs in recognizing it as a crucial and perhaps the most important - step in constructing a basis for thought that will enable scénarios to be built. Without careful analysis of actors' moves, scénarios will lack relevance and cohérence. On the other hand, the same people lament the notable lack of a systematic tool for analysing actors' behaviour. This lack is ail the more remarkable in that anaSysis of actors' behaviour is so often preceded by a radier clumsy structural analysis, using tools (the MICMAC method) to help identify the key variables and ask the right questions - in other words., to improve the pertinence of the thought process. We recall that this is a matter of focusing on those actors who directly or indirectly control the key variables identified by the structural analysis. We then construct an 'actors' stratégies' table, presented in the form of a square matrix (actors X actors) in which: • each diagonal cell contains the aims and objectives of each actor, in so far as thèse can be identified; • the other cells contain the means of action which each actor may use against the others in order to achieve its aims. Fiiling in this table is a group discussion activity, sharing the information gathered on each actor and its relationships with the others. This information on actors' behaviour can be collected or compiemented by conversations with experts who are représentative of each group of actors. Given that it is generally difficult to ask an actor to reveal his own strategy and his own strengths and weaknesses, it is much easier to get him to talk about the other actors. By sifu'ng through sets of partially true information, a more or less cohérent picture of the whole situation émerges. It is often said that it would be good to take advantage of information derived from game theory., in order to make intelligent use of the nearcomplete data coilected in the actors' strategy tables. We support this view, and hâve no doubt that one day young researchers will propose significant ways forward. In the meantime, to our knowledge, the available tools hâve hardly developed at ail over the past few years. 106

Understanding the actors'stratégies: the MACTOR™ method

In 1985, however, we outlined a path which seemed to us promising, and by taking up this idea once again we hâve now created and developed the MACTOR® method (Matrix of Alliances and Conflicts: Tactics, Objectives and Recommendations). Our aim is to create an analyticai toot which will allow us to make better use of the informational added value contained in actors' stratégies tables. Although the 'game theory' patli still appears to be of interest, we did not pursue it rigorously in creating MACTOR®. Others will certainly do so, but we would suggest that they bear in mind the following recommendation: develop tools that are sufficiently simple to be appropriabie (understandable) by the users and which lend themselves easily to multiple and varied applications. Analysis of actors' moves, as we propose with MACTOR®, proceeds in ttie foliowing six stages: 1. Note down each actor's plans, motivations, constraints and means of action (construct the 'actors' strategy' table). 2. Identify the stratégie issues and objectives associated with thèse battlefields. 3. Position each actor on each battlefield and note the convergences and divergences. 4. Rank the objectives for each actor and assess possible tactics (interaction of possible convergences and divergences) in terms of their objective priorities. 5. Evaluate the relationships of power and formulate stratégie recommendations for each actor, in keeping with the actor's objective priorities and available resources. 6. Raise key questions about the future - i.e. formulate hypothèses regarding the trends, events and discontinuities which will characterize the évolution of the balance of power between actors. It is around thèse key questions, and hypothèses as to their answers, that the scénarios will be constructed. The added value created by the MACTOR® method dérives primarily from stages 3 (positioning actors in relation to their objectives); 4 (tactics for possible alliances and conflicts); and 5 (stratégie recommendations). In future, more attention will be devoted to thèse stages, for until now we hâve too often passed rather quickly from stages 1 (constructing actors' strategy tables) and 2 (stratégie issues) to stage 6 (key questions for the future). How do we conduct this analysis of actors' behaviour in six stages? What exactly does the MACTOR® method consist of? To answer thèse questions, we hâve once again chosen to illustrate the method with an example which relies on material coltected while 107

Undentanding iha aaors'stratégies: the MACTOR* method

carrying out several prospective studies in the field of air transportation. Most of thèse studies were carried out in the 1970s (at the time when we ran SEMA prospective), for Aérospatiale, for the Directorate General of Civil Aviation (DGCA) in France, and especially for the Paris Airport Authority. An example like this has not dated - it is now that it is most valuable, for we can check whether or not the conjectures made about the future (which is now the présent) were well founded. Moreover, more récent analyses of actors' moves, relating to futures yet to arrive, are almost systematically confidential. The example of air transportation is currently one of the onty ones that can be made public. Other examples, relating to the Post Office or other iîrms, may perhaps émerge from their wraps of confidentiality after a longer prescribed period. The directors of the Paris Airport Plan hâve confirmed to us that this example of 'retrofutures' retains a certain topicality. Furthermore, a new study of actors' moves in air transportation up to 2010, using the MACTOR® method, has been set up; naturally its results cannot be published, whether they are relevant or not. If an actor reveals to others the nature of the questions he is asking himself, and the way in which he is formulating them, he has already said too much about his strategy . . . unless of course part of his strategy is to use the effect of declaring his hand, as in poker.

1. Constructing the actors' strategy table We are focusing, then, on the behaviour of actors in air transportation in the Paris région up to 1990, as analysed in 1978. The first question concerns the number of actors to take into account. Should we consider the airline companies as a single actor, or should we subdivide them according to a particular characteristic (size, légal status, nationality, etc.)? Similarly, the state is generally a polymorphous actor - there is the DGCA, but also the Ministry of Finance, the government, and so on. Thèse actors, which together make up the state, differ in their objectives, their behaviour, and their criteria for decision-making. A complète analysis would hâve to integrate other actors such as the trade associations, the European institutions, and the international air transportation organizations. One could thus multiply the number of actors at will, and almost inevitably run the risk of making analysis of the System impossible. Expérience shows that a total of ten to twenty actors constitutes a realistic and operational compromise. 108

Constructing tlie actors' siraiegy table

FIG . 17. MACTOR® method: séquence of stages.

The actors' strategy table (based on Godet, 1977) is constructed in a square matrix (actors X actors) (Table 6)3 which we hâve redravvn from memory. The cells on the main diagonal are generaliy thc fullest, for in 109

Undemanding ihe acton'stratégies: the MACTOR» mctliod

TABLE 6 Action of Manufacturera

Action on manufacturera Objïctivï: to survive and avoid crisis Pmblems: Plan for higher • performance aircraft • specitic noise and fuel consumption standards to mect Menus; Association beiween consiruciors

Action on state

Pressure on airlines lo purchase new aircralï

Exercise 'blackmail' in regard to jobs Demand finance for new projects

Diversification of needs and aircraft Standarcliïation of ihe flcet for each constructor Avaiiability of entire range No signifkant lechnological progress

Military orders Diversification of activities

Airlines

Action on airlincs

Demand aircraft better suitcd to their needs

Objectivé: To mainmin market sliarc

Dominant crilcrion: Cosi per passenger mile per ton effectively transporicd

Pyoblmis: Financial invesimcni and salaries To maintain high frtquency and occupancy

Relucuince to use large aircrnft

AL'tws: Co-opcKitioii between

Scek protection from compclilion in the form of discriminaiory rights in relation to longhaul traffic

airiines (ATLAS) Incrcascd use of branches Standardisation and opcraiing flexibility of the flecl Development of freight Concentration at the teriiary levé! (feeder îines) l ' o proictt ihe Prateciionism nation;]] aeronautica! Presstire on airlines to industr>' purchase Mercurj' Sintt

(VSiiitar\p and civil aircrni't order Pinuncc for new projecis

Airbus wiih financial aid Siate protects airlincs provided they develop and improve their section

Approaches to forcign govcrnmenis Appeal 10 privjite finance

no

Objective: Prestige and a French présence in the world Probkms:

Unemployment Inflation

Mains: Sustained growth

Identifyhig the stratégie issues and associated objectives

thèse cells we are seuing clown in black and white each actor's identity card. In contrast, many of the other cells (actions of one actor towards another) are almost or totally empty. In order to simplify this account of MACTOR®, we shall consider only six actors: aircraft manufacturers (Al), scheduled airiines (A2), charter airline companies (A3), the state (A4), Paris Airport Authority (A5) and the local résidents' associations (A6). In the case analysed on behalf of Paris Airport in 1990, twelve actors were considered, as well as seven stratégie issues and over thirty associated objectives.

2. Identifying the stratégie issues and associated objectives Through group reading and discussion of the actors' strategy table, the stratégie issues - i.e. the battlefields on which the actors will confront each otlier - are brought to light fairly easily. Hère we concentrate on five stratégie issues regarding which the six actors hâve converging or divergent objectives (convergences or divergences). Thèse five issues concern the fotlowing: El: Définition of aircraft. The aircraft manufacturers want to impose their own new aircraft designs on airline companies and on Paris Airport. For example, Boeing 747s were developed at a time when the existing runways were too short for them. E2: Aircraft market. National aircraft manufacturers rely on the state to develop their share of the national and international market. The other actors under considération are not concerned with this objective. E3: Allocation of tmffic righis. Hère the scheduled airiines rely on the state to curb the aspirations of the charter companies, who favour deregulation. For its part, Paris Airport supports the opening of new lines which would allow an increase in the number of flights to Paris. E4: Market for 'organized' flights. The interests of charter companies regarding the 'organized' flights market are opposed to those of the scheduled airlines. The main concern of Paris Airport is to avoid having to turn traffic away, and from this point of view it is an objective aiiy of the charter companies. E5: Noise pollution and disiurbance naar airports. This issue is at the

crossroads of actors' stratégies, for it involves ail of them. Résidents demand less noisy aircraft, are opposed to the authorization of

m

Undersianding the actors'stratégies: tlie M ACTORS methocl

night fiights, and their concerns are naturaiiy echoed by the state (résidents are aiso voters). Aircraft manufacturers represent an objective ally of the résidents, in that more restrictive noise-conîrol standards could lead to the abandonment of old aircraft in favour of new, less noisy craft. Scheduled or charter airlines and Paris Airport are naturaiiy opposed to anything which couid curb traffic. Each of thèse stratégie issues (battlefields) can be presented in the form of one or more précise objectives over which actors are in convergence^ in divergence, or neutral. For ease of exposition, we shall limit our analysis to the battlefields (Ol, 0 2 , O3, 04, 05), which constitute only part of the objectives associated with issues E1-E5. A complète analysis of actors' stratégies would hâve to take ail objectives into account. TABLE 7

Issues (battlefields)

Associated objectives

El

Définition of aircraft

01

E2

Market for aircraft

02

E3

Allocation of traffic rights

03

E4

Organized fiights market

04

E5

Noise pollution and disturbance near airports

05

112

• Impose aircraft spécifications (size performance) • Define aircraft spécifications together • Défend and increase the national manufacturers' market share • Maintain allocation of traffic rights • Partial deregulation • Total deregulation (free opening of new lines) • Develop 'organized flights' • Control 'organized flights' • Avoid turning traffic away • Regulate and reinforce noise standards

Posilianing each actor in relation to the stratégie objectives (signed position matrix) .

3. Positioning each actor in relation to the stratégie objectives (signed position matrix) The relationships between the actors on each battlefield can be represented in the form of a diagram of possible convergences and divergences. Of course, in order to understand the stratégie situation as a whole, it is necessary to construct ail the diagrams of possible convergences and divergences associated with each stratégie objective} as well as diagrams of corresponding resources. We soon see that stratégies of convergence and divergence betvveen actors vary from one objective to another. In order to maintain cohérence, there can be no question of flghting against a certain actor on one battlefietd while counung on his support for another, and vice versa. For any given actor, the question is therefore to identify and evaluate possible stratégie options and a cohérent sélection of objectives and alliances. Visual comparison of the diagrams of convergences and divergences is not easy; however, a matrix représentation (MAO Matrix of Actors and Objectives) enables ail thèse diagrams to be summarized in a single table (Table 8). In Figure 18 we see the positioning of actors towards the objective of regulating and reinforcing noise controi standards.

Objective: régulait' and reinforce noise controi standards + = In favourof objective - = Opposée! to objective Fi G. 18. 113

Umierstanditig the actors' stratégies: the MACTOR*1 niethod

TABLÏÎ

8 . MAO: signed matrix of positions (actors X objectives) 01

O2

03

04

O5

Total +

Total -

0

0

+1

+3

0

+i

-1

-1

+1

-3

+1

-1

+1

-3

+1

0

+1

+3

_j

+1

-1

+1

0 _3

+1

0

Manufacturera

Al

+1 + 1

Scheduied airlines

A2

—î

0

Charter eompanies

A3

-1

0

State

A4

0

+1

Paris Airport

A5

-1

0

Résidents' associations

A6

0

0

0

0

+1

TOTAL +

+2

+2

+2

+3

TOTAL -

0

-2

-1

-3

Ol: Impose aircraft spécifications. O2: Défend and increase national manufacturer»' market share. 0 3 : Maintain allocation of craffic rights. O4: Develop 'organîzed fiights'. O5: Regulate and reinforce noise control standards. T h e iMAO matrix (actors x objectives) (Table 8) is fiiled in as foUows: (+1) Actor ( in favour of objective / (-1) Actor i opposed to objective/ (0) Actor ;" neutral in relation lo objective/ So, for cxample, we find the fifth column represents tlie diagram associatcd with objective 0 5 : regulating and reinforcing noise control standards. Commeiilary Simply examining the positive and négative totals of the iines and columns of the MAO matrix provides a wealth of information. Thus we see, on one hand, that tJie résidents' associations (A6) are only concerned with one objective (noise, A5), white ail the other actors are involved in four out of five battlefields. On the other hand> objective O5, noise pollution and disturbance near airports, is the one which most divides the actors and involves them ail. Defining aircraft spécifications ( O l ) , allocation of traffic rights ( 0 3 ) and to a tesser extent the development of the 'organized' fiights market (O4) are also highly contentious objectives.

114

Ranking the objectives for each actor (valued position maivix) and assessing the range of possible convergences and divergences

4. Ranking the objectives for each actor (valued position matrix) and assessing the range of possible convergences and divergences For each pair of actors it is interesting to note the number of objectives over which they are in convergence or divergence. This can almost be picked out visually from tlie MAO matrix. But for largcr tables incorporating about ten actors and twenty or so objectives, we must make use of a classic property of binary matrix caiculation: by multiplying a matrix by its transposition we obtain the number of factors in common for each pair of Unes in the original matrix (to transpose a matrix ail we hâve to do is to place in columns the factors which were previously in lines). The transposed form of MAO (actors X objectives) is called MOA (objectives x actors). The product of matrices MAO x MOA, respectively (6.5) and (5.6) in format, gives matrix MAA (actors x actors) (6.6) in format (Table 9). In order to be able to distinguish which of die factors common to two actors (two fines of the MAO matrix) correspond to motions in favour of certain objectives (indicated by +1) or opposed to others (indicated by -1), we carry out the matrix caiculation MAO x MOA, applying the following conventions: • ncjj is obtained by the matrix product which retains oniy positive scalar products. This is aîso the number of objectives towards which actors i and j hâve a common attitude, either favourable or unfavourable (number of convergences). • ndjj is obtained by the matrix product which retains only négative scalar products. This is also the number of objectives towards which actors i and j hâve a divergent attitude (number of divergences). The matrix MAA (actors X actors) is therefore made up of ail pairs ncjp ndjj. For example} MAA23 = +2, -2 means that scheduled airlines (A2) and charter companies (A3) take up the same position on two objectives (in mis instance Ol and O5), and are in opposition on two other objectives (03 and O4) (cf. the MAO matrix, Unes 2 and 3). Matrix MAA therefore indicates for each pair of actors the number of objectives on which they are in convergence (ne-) or in divergence j; matrix MAA allows us to obtain two complète diagrams of convergences and divergences. Thèse diagrams are shown below; the thickness of the lines is proportionate to the number of objectives concerned.

115

Understanding the acton'stratégies: the A'IACTOR" method

T A B L E 9. MAA: Matrix of convergences and divergences (actors x actors)

Ai

\ Manufacturer Schcduicd airiines Charter companies

A2

0

Al

\ A2 A3

\ 0

A4

0 2

+2 _2 \

A5

A6

0

4

+2

0

+1

-1

0 0

\

-1

+1

0

0 -1

0 0

-2 \

0

-1

0

-2

-2

4

0 0

0

0 -2

-2

-2

-2

\

-1

_2

Résidents' associations

+1 -1

2

-2

0

-ri 0

Paris Airport

0

X

+1

0

-2 +2

A6

A5

+2

0

_2

\ -2 \ -2

State

A4

A3

+1 0

-1

\ \

0 -1

\

First complète diagram of convergences and divergences

The most striking thing is the strong convergence of interest between Paris Airport Authority and the charter companies, and to a lesser extent the scheduled airlines. We also note the lack of common objectives beiween Paris Airport Authonties (A5) and the state (A4) (at least for the objectives under considération). Scheduled airlines Manufacture l'S

+2

(AI Charter companies

(A6 +4

Résidents' associations

Paris Airport

FIG. 19. First complète diagram of convergences over objectives (ncjp. For their part, die aircraft manufaciurers, the state and the local résidents' associations constitute another group of allies on several objectives. 116

Ranktng the objectives for cach acwr (vahted position malrix) and assessing the range oj possible convergences and divergences

Charter companies

Manufactureras (A1

Résidents' associations

F1G. 2 0. First complète diagram of divergences over objectives (jicljp. Complète diagram of divergences

Some actors are in potential conflict witii almost ail the others over two or three objectives. This is the case for Paris Airport (A5), the scheduled airlines (A2) (the actor most at risk), aircraft manufacturer (Al) and the state (A4). Thèse first completed diagrams remain radier elementary because they only take into account the nuniber of convergences and divergences over objectives. To bring the model closer to reality, it is advisable to introduce two dimensions which hâve so far been omitted: • The hierarchy of objectives, which varies from actor to actor. • The relationship of power between actors. Thèse dimensions also affect the interplay of possible convergences and divergences. Before looking at how to integrate the second of dièse two dimensions, we shall examine the first. In order to take into account each actor's spécifie hierarchy of objectives, it is sufficient, for example, to note the positioning of actors in relation to objectives on a scaîe from -3 to 4-3, according to whether the levé! of opposition or agreement is high, médium or low. The more the actor feels concerned witii an objective which is important for him, the higher the absolute value recorded. We thus obtain a second vaiued position matrix of die MAO type, which we shall cail 2MA0 (Table 10); by muhiplying it by its transposed form we obtain a second MAA-type matrix, called 2MAA (Table 11). 117

Understanding thc actors' stratégies: îhc MACTOR™ meihod

T A B L E 10. 2MAO: valued positions matrix (actorsx objectives) Ol

02

O3

O4

O5

Manufacturer

Al

2.0

3.0

0.0

0.0

1.0

Scheduled airlines

A2

-2.0

0.0

3.0

-1.0

-3.0

Charter companies

A3

-1.0

0.0

-3.0

3.0

-2.0

State

A4

0.0

3.0

2.0

0.0

1.0

Paris Airport

A5

-1.0

0.0

-2.0

2.0

-2.0

Résidents' associations

A6

0.0

0.0

0.0

0.0

3.0

T A B L E 1 1. 2MAA: valued matrix of convergences and divergences (actors x actors) A!

\ Manufacturera

Al

A2

\

Scheduled airlines

A2

0 -4

Charter companies

A3

+4

0 -3

State

A4

0

Paris Airporc

A5

0

Résidents' associations A6

\ 0

-4

-2 +8

+4 -4

-3

0

+2 0

-4

-3

0

\ \

0 -3

0

0 -2.5

+2 0 0 -3.5 0 -2.5

0 -3.5 \

+2 0 0 -2.5

0

-4 +8

0

\ -5

+2.5

+4

-2

-5

\

-3 +4

+2.5

+4

\

+2

0 0

-3

-4

A6

A5

+4

0

0

\

A4

A3

0 -2.5 \

The matrix 2MAA is made up of ail pairs of valued convergences (2c;y) and divergences (2d-p. Each élément is obtained as the average intensity (in absolute values) of, respectively, convergences and divergences on objectives. Example: Valued convergence between A2 and A3

US

Evaluaiing the balance ofpower andformitlating stratégie recammendations (valited mairix of position with poioer coefficients)

Valued divergence between A2 and A3 2d23

3

+

-3

+

-1

+ +3

= 5

By définition convergences are noted with the positive sign and divergences with the négative sign. We can thus construct a second version of the complète diagrams of possible convergences and divergences, which in the event does not differ noticeably from the first (which is why thèse second diagrams are not presented hère), except on one point. Between the first and second diagram of conflicts we note an increased degree of antagonism between the scheduled airlines and the charter companies. This dérives from the total opposition of thèse two actors over allocation of traffic rights. Of course, if we had chosen a différent scale for measuring the importance of objectives the results would perhaps hâve developed more noticeably from one diagram to the other. The interptay of possible convergences and divergences does not dépend solely on each actor's hierarchy of objectives, but also on me ability of an actor to impose its priorities on others - that is3 on relationships of power.

5. Evaluating the balance ofpower and formulating stratégie recommendations (valued matrix of position with power coefficients) If we place ourselves in the position of an actor, for example Paris Airport, we see that this actor is in potentiai divergence with almost ail the others over a given objective, while at the same time it may form convergences over other objectives. A cohérent strategy of objectives vvill therefore hâve to impose certain objective priorities. Conversely, defining objective priorities obliges one to formuiate convergence policies. Let us develop the example. Paris Airport has every interest in joining forces with the charter companies (A3) and the scheduled airlines (A2) if it wishes to fight for aircraft spécifications which respond more closely to its constraints (O2), and to oppose new standards and régulations on noise pollution and disturbance near airports (05). This being the case, Paris Airport, out of concern for cohérence, should place 119

Undersiatiding the actors'stratégies: the MACTOR* metliod

on the back burner those issues in which its own interests diverge from those of the scheduled airlines, i.e. traffic rights allocation (O3) and the development of the organized flights market (04). This is particulaiiy important as it is precisely over thèse two objectives (O3) and (04) that the scheduled airlines' interests are opposed to those of the charter companies. For Paris Airport this tactic can only be put into practice if its potential allies., the airline companies., also pursue the same tactic. In reality, everything dépends on how objectives are ranked, which differs from actor to actor, and we should probably expect open conflict between the scheduled and charter airline companies over traffic rights and the 'organized' flights market. Thèse objectives are also very important for Paris Airport, which in principle should first make a bid for convergence with the charter companies. However, merely being in divergence with an actor is not sufficient actively to oppose it - also required are the direct or indirect means to oppose it. The tactical sélection of convergences and divergences is necessarily dépendent on thèse means. Sometimes it is even the existence of a favourable balance of power which sparks off conflict. It is therefore useful to guide one's tactical choice by analysing relationships of power through two matrices - the matrix of means of direct action (MDA) (Table 12) and the matrix of means of indirect action (MIA) (Table 13). The first matrix., MDA, is simply a table (actors X actors) in which the potential influence of one actor over another is recorded on a scale from 0 to 3 (none, weak, average, stxong) - one could use other catégories. This first table aiready reveals relationships of power; simply reading the totals for each line and each column reveals that the state is by far the most influential actor in the System, while at the same time it is one of the most susceptible to the influence of others. In contrast, the charter companies are the least well equipped to achieve their objectives, and are also among those actors most subject to pressure from others. But in iooking at relationships of power, we cannot restrict ourselves simply to means of direct action: an actor can influence another via the intermediary of a third actor. It is therefore useful to examine matrix MIA - means of indirect (second order) action - obtained simply by multiplying matrix MDA by itself (MIA = MDA X MDA) (according to the principle of the MICMAC® metliod). By doing this, we discover that the local résidents' associations are in a stronger position of power than one would hâve thought a priori (ranked second in terms of total indirect influence over actors' moves). This is thanks to their direct influence on the state, the most powerful actor in the System. The charter companies' position of power seems 120

Evalualing tlie balance of power and fonnulating stratégie rocommendaliom (vahied matrix of position zvith power coefficients)

even less favourable than before (they hâve very weak influence, and are highly sensitive to pressure, particularly from the state and the scheduled airlines). T A B L E 12. Apparent reiationship of power: matrix ofmeans of direct action. Total influence

Total influence

Al

A2

A3

A4

A5

A6

S

Manufacturas

Al

0

1

1

3

0

2

7

Scheduled airlines

A2

2

0

3

2

1

1

9

Charter companies

A3

1

2

0

1

1

0

5

State

A4

2

3

3

0

3

2

13

Paris Airport

A5

0

2

3

1

0

2

8

Résidents' associations

A6

0

1

1

3

2

0

7

Total dependence

S

5

9

11

10

7

7

TABLE

1 3 . Real relationships of power: matrix ofmeans of indirect action MIA = MDA x MDA. Total indirect influence

TûIal

indirect influence

Al

A2

A3

A4

A5

A6

M;

Manufactureras

Al

9

13

14

9

15

7

58

Scheduled airlines

A2

7

17

12

13

il

10

53

Charter companies

A3

6

6

13

8

5

8

33

State

A4

9

16

22

24

10

13

70

Paris Airport

A5

9

11

11

13

12

4

48

Résidents' associations

A6

9

15

18

5

il

11

58

40

61

77

48

52

42

Total dependence

As for the Paris Airport Authority, it is in an average position of power in relation to the System as a whole. Its capacity for indirect action over the local résidents' associations is much weaker than its potential for direct action. On the other hand, Paris Airport has significant leverage for exerting indirect pressure over the state vvhile almost totaily 121

Undersianding tiie actors'stratégies: the MACTOR* method

lacking means of direct action. We also observe that the aîrline manufacturer can, if necessary, exert strong indirect pressure over Paris Airport, probably via the intermediary of the state. As we can see, reading the MDA and MIA matrices is a fruitful exercise. Looking at 1978 from the standpoint of 1990, in tlie final analysis Paris Airport Authority had practically no interest in allying itself too openly with the charter airlines, because thèse were the weakest link in the overali balance of power. So it should corne as no surprise that in 1990, at the height of a euphorie growth in air transportation Worldwide, the European charter companîes were 'on their last legs'., to quote Aviation Internationale magazine (No. 996, 15 December 1989). What they should do is to air their common interests with the scheduted airlines, while putting pressure on the state for a degree of liberalization of traffic rights. Naturally, our exampie is oversimplified, and it would be unreasonable to expect to make any définitive stratégie recommendations based on it. We should aiso recall that everything dépends on how each actor prioritizes its own objectives in terms of tlie balance of power. We hâve seen that it was possible to take account of each actor's hierarchy of objectives through the matrix of positions (2MAO). To say that one actor is twice as influentiat as another in tlie overail balance of power is impficitly to ascribe twice the power to this actor's influence over objectives. Its relationships of power between actors are characterized by r- coefficients; it is tlien sufficient to weight the lines of the valued position matrix by thèse coefficients. Thus we pass from matrix 2iVlAO to matrix 3MA0, tlie valued position matrix, weighted by relationships of power. By multiplying 3iMA0 by its transposed form we obtain a third matrix of convergences and divergences. How should we define thèse ?y indîcators of relationships of power? The first idea that cornes to mind is to consider tlie M;- indirect influences in the matrix of real relationship of power. The measure of relative indirect influence (M;-/SMy) gives a good indicaior of die power of one actor over the others. However, with identical relative influence, one actor will be in a better position of power tlian another if its overall dependence is lower. So we must balance the preceding coefficient M.(7XMj with an inverse function of dependence M ; 7(Mj + D2-). With?7 =

' *

122

x ~ ,' _

Evaluating ihc balance of power and foninilating stratégie recommendatioiis (valued matrix of position vihh power coefficients)

if Dj- dependence is zéro, ?y = Mj/ibAf, if Dj dependence is strong in relation to influence, then the ?y relationship of power will be much weaker than the simple reiationship M^/XM,-. Moreover, in order îo facilitate understanding and calculation, we suggest considering

Starting with tiie matrix of real relationships of power, in our example, we obtain the balance of power coefficient for each actor: r,* = 1.23, r2* = 0.88, r3* = 0.36, r,* = S.49, r5* = 0.83, r6* =1.21 The sum of thèse coefficients is equat to six. If ail the actors had the same relationship of power, ail the r;- would be equal to one. In this example the hierarchy of power relations coefficients is exactly the same as chat of overall influence. It is the state (A4) which has the most favourable position of power, followed by the manufacturers (Al) and the résidents (A6). At the other end of the spectrum, the charter companies (A3) represent die weakest link in the game. We pass from the valued matrix of position 2MAO to the matrix of vaiued positions balanced by relationships of power 3MAO by multiplying each Une of 2MA0 by the rj coefficient (Table 14). TABLE;

14. 3MAO Ol

02

03

04

Al

2.5

3.7

0.0

0.0

1.2

A2

-1.8

0.0

2.7

-0.9

-2.7

A3

-0.4

0.0

-1.1

1.1

-0.7

A4

0.0

4.5

3.0

0.0

1.5

A5

-0.8

0.0

-1.7

1.7

-1.7

A6

0.0

0.0

0.0

0.0

3.6

123

05

Understanding ihe aciors' stratégies: the MACTOR*3 metliod

We thus obtain a third matrix of convergences and divergences (3MAA) balanced by relations of power (Tabie 15). TABLE

15. 3MAA

Manufactureras Scheduled airlines Charter companies State Paris Airport

A2

Al

\ Al

\

0 -4.1

\

A3

A4

A5

A6

0

2.4

5.4 0 -2.4 0

-3 1

0

\ 2.7 0 0 2.8 3.5 -2 8 -2 .1 -3 4 -4 1 \ -3 1 A3 0 0 27 4.5 0 -2.8 \ 0 -2.2 -3 .1 -2 4 \ +2 8 0 +2.6 A4 +5.4 0 \ 0 -2.1 -3 1 -3 9 0 \ A2

A5

Résidents' associations A6

0 4.5 35 -3.4 0 -3.1 2.4

0 -3.1

0

0 \ -3 .9

2.6 0 _2 2 0

\

0 \ -2.6

0 -2 6

\

Commentary Betvveen the ftrst and the third complète diagrams of convergences (Figs. 19 and 21), certain deveiopments are worth pointing out. The state manufacturers' convergence over objectives becomes noticeably stronger and appears twice as important as the state résidents' convergence over objectives, which initially appeared comparable. The convergence over objectives betvveen Paris Airport and the scheduled airlines and charter companies is confirmed as much stronger than the convergence of interests between the companies (scheduled and charter), and is probably a card to be played by Paris Airport. Comparison of the diagrams of divergences over objectives (Figs. 20 and 22) allows us to note certain remarkable changes in the actors' strategy when hierarchy of objectives and balance of power are taken into account. Thus, for exampfe, the opposition of interests between the scheduled airlines and the manufacturers seems to be twice as signincant as that between the charter companies and the manufacturers or scheduled airlines, which initially appeared comparable. In the same way, the conflicting objectives of the state and Paris Airport become much more critical than the conflict of objectives between the state and the airline companies. 124

Evaluatittg the balance of power andfortnulating stratégie recommendations (valued matrix of position viith power coefficients)

Schedulcd airlincs Manufacturers

2.7

Charter companies

(A6, 4.5

Résidents* associations

Paris AirporE

F i G . 2 1. Third complète diagram of convergences over objectives. -2.4 Schedulcd airlincs

Charter companics

Manufacturers

Résidents' associations

FiG. 22. Third complète diagram of divergences over objectives. From the above, it is not unreasonable to conclude that tlie state should support the manufacturers in their struggie for market share, and should strengthen régulations and standards vvhich favour the development of new aircraft. Paris Airport, which is subject to the powerful protection of the state, must above ail rely on the scheduled airlines to exert pressure on the state, for the charter companies are in a much less favourable position of power. In doing so, Paris Airport should logically tone down its support for development of charter flights, because the scheduled airlines are opposed to this. 125

Understanding the aaars' stratégies: the MACTOR*1 method

In the example \ve hâve considered, taking into account hiérarchies of objectives and relationships of power did not cause major upsets to the fîrst anaîysis. Other scales of notation would probably provide a clearer contrast in results. Other indicators of power relations could also be tested, for example, by relying on the matrix of net balance of power between actors (derived from MIA matrix through the différence between a;y and a-). TABLE

16 Al

A2

A3

A4

A5

A6

6

8

0

6

-2

+ 18

6

-3

0

-5

-8

-8

-6

-10

-38

-3

8

+ 16

-7

-4

Manufacturers

Al

Scheduled airlincs

A2

-6

Charter companies

A3

-8

-6

State

A4

0

3

8

Paris Airport

A5

-6

0

6

3

Résidents' associations

A6

2

5

10

-8

7

+8

+38

-16

+4

\ \

-18

\ \

\

+ 16

\ -16

Three actors (manufacturers, state and résidents) are overall in a very favourable position of power in relation to the group comprising Paris Airport and the scheduled airlines, which are overall neutralized; and the charter companies are completely dominated, i.e. in an unfavourable position of power in relation to ail the other actors. Paris Airport is better placed in relation to the state than one wouîd think a priori. The potential direat cornes primarily from the manufacturers and the résidents. We shali not pursue this simpiified example any further. For the most complex cases - around fifteen actors or ten stratégie issues and about forty objectives - one certainly has to break down the problem by studying each battleneld separately. This modular utilization of the MACTOR® method implies, of course, the never-ending task of developing a cohérent overall picture with each further addition.

6. Key questions for the future The évolution of relationships of power between the actors can be presented in the form of hypothèses which may or may not be realized 126

Key questions jar ihe future

within the time horizon under considération. Thèse hypothèses are concerned with trends, as much as with events or discontinuities. The subséquent application of the scénarios method consists of using expert methods to reduce precisely this uncertainty over hypothèses concerning futures hypothèses deriving from the actor strategy analysis. We believe that MACTOR® will disseminate rapidly, as it is a simple and appropriate tool which will lead to a better understanding of tlie actors' games and power relationships, The air transport example developed hère was rather illustrative, We assume that the use of MACTOR® by Paris Atrport Authority and by the French Electricity Authority to face the new European context proves that MACTOR® certainly has a promising future.

127

6. Reducing uncertainty: expert consensus methods

In the same way that the past can be summarized by a séries of significant events, possible futures can be identified by a list of hypothèses which, for example, demonstrate the persistence of a trend} its disruption or the developmental genesis of a new trend. In concrète terms, thèse hypothèses relate to the key variables and the balance of povver between the actors as analysed when the base was constructed. Scanning possible futures is done through morphological analysis. The implementation or otherwise of the hypothèses within a given time-frame is marked by a degree of uncertainty which can be reduced with the aid of subjective probabilities provided by experts. In fact, when one looks to the future, personal judgement is often the oniy way to account for events which might occur (there are no statistics of the future). Expert methods prove useful for reducing uncertainty and also for comparing the views of one group with those of others (bellweather function) and, at the same time, for becoming aware of the scope and span of opinions held. The Delphi mediod, developed in the 1950s, is the best known expert method. It suffers from a number of weaknesses, however, notably in its failure to account for die interdependence of the questions asked. A new method then appeared at the end of the 1960s in the United States and at the beginning of the 1970s in Europe: crossimpact matrix. Foliowing an inventory and historical review of thèse tools, we provide as an example a method which we developed in 1974: the SMIC method.

129

Reducing uucertaïnty: expert consensus metltods

1. Scanning possible futures: morphological analysis' This is a sophisticated title for a very simpie, yet often unappreciated (or forgotten) method which should be reinstated, for it can prove very useful for stimulating tlie imagination, hclping to idenu'fy hitherto unknown products or processes and scanning the fieid of possible future scénarios. The originaior ofjthe method, F. Zwicky, envisaged that this method wouid 'make invention routine', i.e. a commonplace procédure. Zwicky, who was the first to imagine dwarf stars, developed the method in the 1940s while vvorking for the United States Army. According to legend it was through this method that Polaris missiles were first thought of.

1.1.

THE

FIELD

OF

POSSIBLE

FUTURES,

A M O R P H O L O G I C A L SPACE

The principle underlying the method is extremely straightforward. The System or function under study is broken down into subsystems or component parts. Thèse components must be as independent as possible and yet représent the total!ty of the System under study. An aircraft is composed of wing structures, tail units, engines, etc., and each of thèse components could be assembled in différent configurations, for example, one, two, three, or more, wings. A given aircraft will therefore be characterized by the chosen spécifie configuration of the components. There are therefore as many possible technical solutions as there are possible combinations of the configurations. A System of four components, each having four configurations, represents no fewer than 4 x 4 x 4 x 4 = 256 combinatory possibilités. This field of possible combinations is called the 'morphological space'. In the example proposed by E. Jantsch (1967), concerning propulsion units, there were more than ten components, each with two to four possible configurations: thus tlie total number of imaginable solutions was 36,864. Fortunately, constraints exist (technical incompatibility, for example) which make several families of solutions unworkable and thus reduce the morphological space. It is also possible to give préférence to those combinations of configurations which appear I. To Icarn about morphological analysis, sec E. Jantsch (1967) and R. Saint-Pau! and P. F. Téniêre-Buchot (1974). UnfominaEely thèse two excellent studiesiiavcbeenout of print for severai years now. See also F. Zwicky (1947). 330

Scarining possible futures; morphological analysis

more propitious than others, in terms of such criteria as development costs, safety factors, feasibility, etc. It is useful to compare the existing with the total morphological space: the harvest of new icleas will be ail the richer in that the potential space is unknown (the ratio of the number of known solutions to the number of possible solutions is particuiarly interesting in this respect). Areas of application are numerous and extend to many innovation and research procédures. R. Saint-Paul and P. E Ténière-Buchot (1974) remind us of the amusing example of the shaving function (Table 17). TABLE

17.

Shaving

Energy sources Shaving agents Dynamics

o

0

o

Electrical

Chemical

Manual

o

Mcchanical

0

o

o

O

0

Heat

Elcctricity Blade

Chemical

Bactcria

0

o

0

Circular

Linear

S ta tic

This morphological space contains 4 x 5 x 3 = 60 possibilités. We recognize the currently used methods in the above table: the linear dynamics manual blade and the circular blade electric shaver. The authors cite some solutions worth studying, such as 'sélective burning ofthe beard hair, or bactcrial ingestion ofthe beard by siatic application of a crcam'.

Curiously, morphological analysis has been used primarily in technoiogical forecasting and far less in gênerai futures studies. Hovvever, it lends itself well to the construction of scénarios. Let us consider a globat System broken down into components: démographie., économie, social and organizational factors, with a certain number of possible states (hypothèses or configurations) for each of thèse components (Table 18). TABLE

18. Configurations (hypothèses)

Components (dimensions) Démographie factors Economie factors Technotogica! factors Social factors In essence, a scénario is no more than a path combining a given configuration for each component. The morphological space defines the range of possible scénarios very precisely. It is probably the fear of 131

Retlucing uncertainty: expert consensus methods

drowning in the huge number of possible combinations that has curbed the utilization of morphological analysis for the construction of scénarios. We propose below a methodologicai approach which sidesteps this obstacle and enables one to make the most of the analysis.

1 . 2 . PROBLEMS AND LIM ITATIONS

Specifically, the utilization of morphological analysis poses several problems relating to the question of comprehensiveness and to the limitations and illusions inhérent in the number of combinations. First]y, the choice of components is particularly critical and requires considérable thought. By multiplying the number of components and configurations, the System expands very quickly, to such an extent that analysis rapidly becomes practically impossible; on the other hand, vvith too few components the System is impoverished. Hence the requirement that there should be a compromise value. One must be careful to ensure that the components (dimensions) are independent and that they are not confused with the configurations (hypothèses). Secondly, scanning ail possible and imaginable solutions may give the illusion that ail combinatory possibilities hâve been expiored, whereas in reality the field has no définitive limitations but is simply evolving witli time. Should one leave out some component or configuration which is essential for the future, one risks mtssing a whole facet of the field of possibilities (had such a study been applied to timepieces at the start of tlie 1950s or 1960s, everything woutd hâve been imagined except me quartz watch). Lastly, the user is very quickly submerged by me number of combinations and simply articulating the possible solutions becomes physically impossible once they exceed several hundred in number (we saw above that for a System with four components and four configurations each, there were already 256 possible combinations!). In such conditions, how can we identify the useful morphological subspace?

1.3.

THE

USEE-UL M O R P H O L O G I C A L S U B S P A C E

Reducing the morphological space is necessary because it is impossible for the human mînd to cover, step by step, the whole field of possible solutions generated by the combinations; it is also désirable since there is no point in identifying solutions which in any case will be rejected once certain sélection criteria (technical, économie, etc.) are taken into Î32

Scan/ring possible futures: morphological analysis

considération. Choices are therefore imperative to identify the key and secondary components which relate to thèse criteria. In a récent example, concerning a future, we suggested the following procédure: • First identify the économie, technicai and stratégie sélection criteria which can be used after the morphological analysis, to assess and sélect the best solutions among the total number of possible solutions (the morphological space). • Then identify those components which are deemed crucial according to the above criteria and classify thèse components in terms of criteria weighted differently according to the policy décisions adopted. • Initially restrict exploration of the morphological space to the key components identified in this way. • Finally introduce constraints of exclusion or préférence (thèse combinations being excluded). In fact, many technicai solutions are meaningless or irrelevant, either because of their intrinsic incompatibility (impossible combinatory associations) or once certain criteria (for example cost, competitiveness) are taken into account. This procédure enabled us to give priority to examining four main components out of nine under considération. The initial morphological space, comprising 15,552 possibilités, was thus reduced to a useful subspace containing about 100 solutions (i.e. a réduction factor of 150). Taking the technicai incompatibilities into account enabled us to eliminate more than half of thèse. The remaining half were subjected to individual assessment, each criterion being assigned a weighting value according to its importance in the policies under investigation. Thus, for each policy (set of criteria weightings) we were able to rank the fifty or so remaining solutions. A comparison of the various rankings then !ed to the identification of a hard core of twenty solutions, comprising the best solutions in terms of ail criteria and policies, as weli as certain solutions which were outstanding in terms of certain spécifie criteria, and which for this reason were supported by a particular member of the working group. The twenty solutions at the hard core of the useful morphological space were then grouped into families, i.e. in terms of their relative degree of kinship (identicai solutions, give or take one or two configurations). Each of thèse twenty solutions was then subjected to detaiîed analysis and évaluation, specifying tliose configurations relating to the rive secondary components. From this experiment, we note two important lessons for the implementation of morphological analysis: • It forces us to think in a structured way about which components and configurations should be taken into considération and enables a systematic scan through the range of possibilities. 133

Reducuig uncertaituy: expert consensus melhods

• Aithough the sum of possible combinations should not give the illusion of comprehensiveness, it shouid not be ailowed to paralyse the thought process. It is fairly easy to reduce the analysis to a useful morphological subspace (10, 100 or 1,000 times smaller). To do so, one only has to introduce sélection criteria (with die help of Multipol, a multicriteria method) and, for example, constraints of exclusion or préférence. The method developed above is now integrated into a microcomputer software package (Morphol®). Morphological analysis can no longer stand accused of being difficuk to implement. It is our feeling that this method will be the subject of renewed futures studies. Theory can only be crédible when confronted with the test of a practical application.

2. An inventory of expert-consensus methods Methods involving expert consultation can be ranged under three headings: 'Delphi', the voting machines and 'cross-impact'. 2.1.

THE DELPHI

METHOD

The Delphi method2 involves 'The systematic use of the intuitive évaluation of a group of experts' (N. Dalkey and O. Helmer (1963)). According to O. Helmer, one of the founders of the technique, 'Delphi' was developed from the middle of the 1950s to meet the needs of the United States Army.( According to some authors, it has seen several thousand applications in the United States and several hundreds in Europe. Thèse esùmates should be viewed widi some caution inasmuch as Helmer himself refuses to propose figures. The Delphi method interrogates experts by means of successive questionnaires in order to bring out convergence of opinions and to identify clearly a possible consensus. The inquiry is carried out by mail and is anonymous in order to adjust for 'leading opinions'. The questions, by way of an example, deal with die probabilities of impfementation of a certain hypothesis or event. The quality of the results is heavily dépendent upon the care taken when dravving up the questionnaire and in the choice of the experts. To be relevant, the method présupposes, on die one hand, diat 2. The word Delphi commémorâtes the town of Delphos, and its famous oracle. 3. Interview given in Scpiember 1984 at Carmcl, California, where Dr HeJrrter is in retirement.

134

An invcnlory of expert-consensus methods

genuine experts are consulted, i.e. people who are really able to answer die questions, and on the other hand that the view of a group of experts is superior to that of a single expert. The Delphi method compounds a séries of mailed questionnaires, their analysis and interprétation. As G. Ducos (1983) reminds us, each expert is offered 'compensation and asked to answer only those questions for which ne regards himself as most compétent, or better still, to evaluate his own level of compétence with regard to each question*. The aim of successive questionnaires is, as R. Saint-Paul and P. F Ténière-Buchot (1974) note, 'to reduce the interquartile space whiie specifying the mean value'. To illustrate the method we will take an example which is very similar to that in their book. This Delphi inquiry covered a whoie range of économie questions. We will follow up one of them: In what year will the gross individual income in France be double that for 1984? The first questionnaire aims at idenrifying the mean value and the interquartile space.4 The distribution of the responses to the first questionnaire (see bar chart) shows that the mean is located in 2000 and the interquartile space (Ql, Q3) is between 1997 and 2005 (see Fig. 23). The second questionnaire is aimed at reducing the interval (Ql, Q3) by F I G . 2 3 . Question: In what year wifl the gross individual income in France be double that for 1984 (in constant francs)? (After Saint-Paul and Ténière-Buchot, 1974.)

askîng for justifications of the extrême values. Each expert is informed of the first round results, is asked to supply a new response and to justify where itlies outside (Ql, Q3): Response Former

New

Justification: when replying outside range (Q1.Q3)

4. Hère the mean value (second quartile) is the year that 50 per cent of the experts think that the doubling will occur 'beforc', and 50 per cent think it will take place 'after'. In taking the 2 5 and 75 per cent thrcsho!ds, then 75 and 2 5 per cent, wc défi ne respectivcly the first quartile (Ql) and the third quartile (Q3). The interquartile range is the interval (Ql, Q3). 135

Reditcing uncenainty: expert consensus methods

It is easier to conform, i.e. to answer in the întervai (Ql, Q3), than to maintain thereby an extrême point of view that has to be justified (if you think like everyone else, no one asks you why). The simulated convergence gives a new reduced interval (Ql, Q3) 1998-2004 and a new mean value 1999. The ihird questionnaire is designed to oppose the two extrêmes by bringing their arguments together. In addition, each expert is asked to criticize those who fail short of Ql (1998) or beyond Q3 (2004). On this occasion the convergence is forced, and even manipulated, since only the extrêmes are opposed, whereas taken together they represent as many replies as are in the interval (Ql, Q3). Also, those at the extrêmes are not asked to criticize the views of those situated in the interquartile space: Arguments 'beforeQl'

New response

Arguments 'after Q3'

Criticism and arguments

Thefourth questionnaire gives the final response: 1999 for the mean value, 1998 for Ql and 2003 for Q3 (Fig. 24). 2010 i -

2005

2000 1997

Qi

1990

Itérations No. 1 No. 2 No. 3 No. 4 (A classical poil resuit) (Deîphi resuit) F I G . 24. The Delphi method, as described, to the extent that there seems to be a convergence of opinions around a central value, provides the right way to prépare a consensus for certain types of decision-making (technological investment with high économie and social risks). However, convergence does not imply cohérence and consensus does not necessarily mean a correct forecast (everyone could be wrong at the same rime). The history of forecasting errors teaches rather that we should distrust prédominant ideas; the correct viewpoint is often to be found in the minority. 136

An inveniory of expert-consensus methods

In short, if the Delphi process appears to be weil adapted to normative applications in its classical form, it could prove more misleading than useful for forecasting applications. There is naturally no reason why one shouîd not alter the method and make Delphi a more usefuî forecasting tool by abstaining from reducing the extrêmes. There is less need to put out severaî rounds of questionnaires, vvhich proves long, costly, painstaking and sometimes risky (since at each itération a certain number of the experts evaporate). For forecasting applications the mini-Delphi technique (see Ducos, 1983) seems more appropriate. Using this technique, firstexperimented with in the early 1970s (see Helmer, 1972), the experts are in the same venue and discuss each question before responding individually. The mini-Delphi process is likely to expérience a vogue in various forms owing to the new possibilities opened up by microcomputers in the use of voting machines. We must mention hère two promising developments, one in France (Regnier's Abacus) and the other in the United States (Consensor).

2.2.

VOTING MACHINES: REGNIER'S AND T H E C O N S E N S O R

ABACUS

Regnier's Abacus (Régnier, 1983, 1989) is a novel method for expert consultation. Its originator, Dr Régnier, a consultant economist specializing in organization and communication methods, believes that the method (successfuily used in the pharmaceutical industry) could now be developed more systematically. The main idea of the Abacus is to start with the three traffic-light colours of green, amber and red, and then to add pale green and pink to allow more shades of opinion. White is a bîank vote, black is an abstention. The colour-coded votes are then placed on a grid like a crossvvord puzzle. Each participant ha s a column and each élément of the problem a row. The mosaic image sketches out a map which guides verbal discussion. Nevertheless, the procédure remains open and anyone can change their colour at any time. The Abacus is not a test but a tool vvhich stimulâtes interaction among individuals. While we can only talk in turns, the Abacus image reveals everyone's opinions on the problem simultaneously. The Abacus is therefore an instrument tending to make communication efficient, for example, in a group where certain individuals are often stifled. What is more, the act of announcing a colour before the discussion expresses an individual's feelings and goes beyond mère rationaïities. 137

Reducing uncertainiy: expert consensus methods

Questions

Light Green green Orange Pink

Red

White Black

1 2 3 4 5 6 7 8 9 10 II 12 13

f/tw to make your choice using ihe grid The colours hâve thc fotlowing meaning: Green, like traffic-light green, is a rather favourable responsc. Red, like trafficiight rcd, is a somewhat unfavourabte reply. Orange, like the traffic-light amber, is the intermediatc colour between reci and green. Lighi green is a shade more positive than orange, but less positive than green but not as unfavourable as red. Thus wc hâve a décision scaie with five steps: five shades: green, light green, orange, pink, red. Moreovcr, a sixth coiour, white (= bîank vote), implies you do not wish to décide on a cotour. Black means you wish to abstain from voting.

Regnier's Abacus was published, with its associated pack, in 1975. It comprises four coloured cubes vvhich slot into a honeycomb tray and has been used in various firms and institutions and in higher éducation (the National Polytechnic Institute of Lorraine, the Strasbourg Institute of Social Psychology, the École Supérieure de Commerce de Paris etc.). The automated Abacus, programmed by microcomputer, is more récent. Image processing, for example the permutation of the rows and columns for sorting, has become instantaneous. This manipulation of the coloured texture allows an easier reading of the pattern in the mosaic. The tool also has the attributes of efficiency, time-saving and simplicity, Learning time is ver y short and the meaning of the image is 138

An inventory of expert-consensus methods

understood internationally. The Abacus reconciles the individual with the group through exchange and it introduces an approach which allows us better to exploit qualitative aspects, with which our society is less familiar than it is with quantitative anaiysis. In its current form Regnier's Abacus makes clever use of familiar colours to exploit collective opinion, yet it possesses certain characteristics spécifie to manual coding that limit its applicability. Consensor, an electronic voting machine developed in the United States by the firm Applied Futures,5 allows a synthesis that is: • immédiate: a histogram shows participants' reactions within a few seconds; • confidential: the anonymous votes of each individual can be transmitted; • scaled: the votes range from 0 to 10; • weighted: the opinions of the participants are weighted according to their respective compétence; • continuous: the expert can change his vote in the course of the discussion. Consensor is portable and allows a group of up to twenty or so experts to express their opinions at a time. Given that tools such as Regnier's Abacus and Consensor are apparently simple to use and inexpensive, why hâve they not yet received the development they merit in Europe? There are apparently numerous applications, both in prospective studies and in décision-support. Consensor would also appear to be useful for collecting information necessary for cross-impact anaiysis and for Regnier's Abacus.

2.3.

CROSS-IMPACT

METHODS

Although the Delphi method enables a fairly good synthesis of opinions to be obtained from which a convergent resuit can be determined, its shortcoming is that it does not allow for interaction between events. On the other hand, the cross-impact method (CIM)'' has the advantage that it takes into account expressed opinion and simultaneous interdependence between questions, providing an interprétation grid which proves cohérent. The cross-impact method is the generic name for a family of techniques which attempt to evaluate changes in the probability of occurrence of a given category of event, follovving a previous occurrence 5. Applied Futures, 22 Greenwich Plaza, Grccnwich, CT 06830, United States. 6. Somc authors, îike G. Ducos (1980), speak of interaction anaiysis. 139

Reducing unccrtainty: expert consensus methods

of such an event. The method starts with a list of events and their associated probabilities; the basic hypothesis of the method is that the individual probabilities account for interactions, but only incompletely. Taking the interdependencies into account allows us to move from a System of unprocessed initial probabilities to a set of net probabilities, i.e. corrected probabilities. The rest of the method consists, first, of analysing the sensitivity of the event analysis System, and, second, of building scénarios. The working out of the scénarios is implemented by highlighting the most probable final images. In practice, if we hâve a System oîNhypothèses (h\,h2 . .. hn), there are 2" final images (interacting hypothèses), that is, ail the scénarios possible for the System. For example, let there be h\, h2, M . . . hn but not }ii: this is one of die 2". If we take into account the order in which the events occur, there are A/! 2n development scénarios. Several methods of cross-impact analysis hâve been put forward. To start with T. J. Gordon (Gordon and Hayward, 1968), interactions are evaluated with impact coefficients ranging from +10 to -10, the transition between rough and final probabilities usually being achieved by fairly sophisticated techniques: the Monte Carlo itération method. Later research, mainly that of J. Kane (1972), fell into the methodological framework set up by T. J. Gordon and continued to mix impact coefficients with rough probabilities in the formulae leading on to final probabiiities. From this viewpoint, N. Dalkey (1972) was an exception: his mode! relied on a conditional probability matrix exisu'ng between every pair of events, to modify the original System of probabilities. This approach, which represented a dennite improvement with respect to CIM as proposed by Gordon and developed by Dalkey, is not very crédible. In effect, the results obtained dépend on the transition formula adopted for calculating the final probabilities. Several formulae hâve been proposed, often a judicious mixture of quadratics, mathematical expectations and coefficients with subjective weighting. In practice none is convincing and, as Dognin and Florentin (1973) demonstrate, there can be as many results as formulae tested for a given single example. Moreover, the method ought to be used to control the cohérence of the estimâtes relative to the classical constraints on the probabilities. In practice, most methods, whatever their degree of complexity, resuit in clearly incohérent probability values, with, for example, results such as P(i) < p(ilj).P(ijnot j), which is not compatible with the relationship P(i) - P(i!j)Pj + P(i\noi j).P(not j), which must always be valid. Many

writers confuse convergence with cohérence: because results are cohérent, that does not mean that the process is necessarily convergent. HO

An inventory oj expert-consensus methods

Two methods, EXPLOR-SIM and SMIC,7 presented in 1974 represented a décisive step forward by insisting on working with homogeneous data (probabilities only) to seek cohérent results and work out scénarios, The first method was developed by researchers at the Battelie Institute, Geneva (Duval et al., 1974) and the second by Duperrin and Godet (1973). Other methods hâve been proposed since, and hâve been the subject of numerous présentations and polemic debates in Futures and Tcchnological Forecasiing and Social Change. Several French researchers

hâve played an active rôle in this development, for example J. Eymard (1975) with a Markovian model or G. Ducos (1980) with the MIP (1) and MIP (2) methods. The debate mainly centres on one essential point, namely: 'What questions may reasonably be put to an expert and how does one dérive cohérence from the answer?' Under certain conditions, an expert can answer with simple conditional probabilities for severai pairs of hypothèses i and j. Unfortunately it is practically impossible for his responses to validate the classic axioms governing probability: summing and product rules (constraints (b) and (c)).K The search for cohérence is ail the more délicate in that it is Iegitimate to présuppose that some answers are more reliable than others (unevenly distributed incohérence). Each method provides a spécifie solution to the problem of cohérence, but none is totally satisfactory. Thus, for example: • EXPLOR-SIM only asks the experts for simple probabilities P(i) and certain conditional probabilities P(ijj) to caiculate P(ijnotj) taking into account the summing rule. Although this practice has the advantage of constructing a matrix cohérent with respect to constraints that are anyway insufficient, it nevertheless has the serious deficiency that it regards the estimate for P(ilj), for example, as fully reliable. • SMIC invites the experts to answer a grid of ail questions P(i)P(ijj), Pftlnotj), and seeks to develop information as cohérent as possible with the initial set of data,, by minimization of a bounded quadratic form. In so doing, however, data with différent degrees of reliability are processed identically.

The SMIC method (French acronym for Cross-impact Systems and Matrices), developed at the Programmes Department of the Atomic Energy Commission (CEA) from 1972 to \ 973, and further developed at SEMA. (b) P(i.j) = P(ijj) PO) = POU) PO); (c) P(i) = POU) Pj + P(i!noi j).P(noi j).

E41

Reducing uncertainty: expert consensus metlwds

The debate is still open, though at a standstill, and the methodological progress of the past few years, with oniy a few exceptions,'' relates to minor refinements (see the excellent summary by O. Helmer, 1981). Finally, on both sides of the Atlantic, certain methods, in spite of their imperfections, hâve demonstxated their usefulness through a significant number of practical applications. Such is the case, at any rate, with the SMIC method, which is described belovv.

3. The SMIC method Among cross-impact methods, the SMIC method offers the advantage of very simple and inexpensive implementation (drawing up a questionnaire), provides rapid results and is generally easy to interpret. The SMIC method consists of intervievving a panel in the most rational and objective way possible. A set of formai interviews could be conducted and an overall measure of agreement obtained, but unless a great deal of time vvere devoted to this, it would be difficuit to see many more than twenty persons. With the metliod used hère, two or three times that number can be questioned. Thus it is possible to exceed the threshold (around thirty peop-le) beyond which past expérience shows mat the law of large numbers opérâtes within a somewhat restricted overall professional environment. In other words, much the same results will be obtained, whether we ask 100 or 150 people. The method usually takes the form of a mailed inquiry, thus allowing the élimination of subjectiveness due to the researcher's présence, since, instead of a verbal opinion, évaluation will be based on replies couched in figures. The inaccuracy of expert answers is reduced: experts are asked to estimate the probability of a hypothesis occurring, on a scale from 1 (very low) to 5 (event highly probable). They are also asked to estimate in the form of conditional probabilities the likelihood of a hypotliesis coming true as a function of the other hypothèses. The value 6 corresponds to the independence of a hypothesis. As a resuit of this, the expert has to revise his assessment several times and, to some estent, he must reveai the implicit cohérence of his reasoning. The reverse side of the coin is the difficulty of representingthe future of a complex phenomenon by a limited number of hypothèses, which runs contrary to the approach in some opinion-gathering methods. 9. Cf. Dr Enzer's ( 1980) INTERAX method at the Ccnter for Futures Research of tlie University of Southern Caiifornia. 142

The SMIC mcihod

Thèse,, however, do not take account of the interdependence among the questions and quite often resuit in contradictory answers. That too can prove an advantage: the method requires the préparation of information and that the issues be considered carefuily before sélection of the key hypothèses. This underlines the importance of structural analysis and the need to understand actors' stratégies in order to identify key variables and formulate stratégie hypothèses. The amount of information gathered in the course of a SMIC inquiry (sixty to eighty questions usuaîly answered by forty to sixty experts) is of the same order as the amount furnished by a conventional survey asking a représentative sample of 1,000 people two or three questions. Thus, the choice lies between canvassing a wide spread of opinion for a small number of questions and an in-depth analysis of the 'global vision' of a small number of experts.

3.1.

THE

P R I N C I P L E AND

T H E ASM

OF

SMIC

An event is a hypothesis coming true at a given time horizon and the set of hypothèses constitutes a référence frame within which there are as many possible states, i.e. final images, as there are combinations of outeomes of the hypothèses. The SMIC method, starting from information supplied by experts, allows a choice to be made from among die 2n possible images to sélect those which merit a more detailed study, allowing for their probability of occurrence. The experts are asked (as a group of individuals) to tender information on: • die list of N hypothèses considered fundamental to the objectives of the study: H=(Hl, H2, . . . HN); • die probabilities of an implementation at a given time horizon: P(i) probability of hypodiesis Hi; • die condidonal probabilides of die hypodieses (paired): P(i/j) probability of i if y occurs; P(ilnotj) probability of i if y does not occur. In practice, the opinions supplied in response to some spécifie questions regarding non-independent hypodieses are not cohérent vvidi respect to the conventional constraints bounding probabilities. Thèse unprocessed opinions must be corrected in such a way that the final results validate the following conditions: (a) O^PQ^ 1 • (b) P(i!j).P(j) = P(i!j).P(i) = P(Lj). (c) P(ilj).P(j) + P(Hnotj).P(noij) = 343

Reduciug uncertainly: expert consensus meihods

The principle of the SMIC method is to adjust the experts' unprocessed opinions in such a way as to obtain cohérent final results (satisfying the normal constraints bounding probabilities) which remain as close as possible to the initial estimâtes. We could envisage optimizing some function of the individual or conditional probabilities subject to the above constraints. However, nonlinearîty of the constraints governing the isolated hypothèses imposes spécifie conditions on the optima; this leads us to concentrate on the combinatory probabilities of hypothèses., i.e. situationa! scénarios. The principle adopted attempts to obtain cohérent final probabilities from the hypothèses"1 which resuit from the combined probabilities of the hypothèses, and which form global, unexpressed but implicit opinions relating to the scénarios: Unprocessed individual and conditional probabiiities Initial data

—»-

Probabilities of images or scénarios Transformation

^

Final individual and conditional probabilities Final, cohérent data

3 . 2 . RELATIONSHIPS E E T W E E N HYPOTHESES AND SCENARIOS: THE SEARCH FOR COHERENCE

The 2N = r possible situations of die System Er comprising N hypothèses, are: El = (M, A2,. . .hi. .. hN) E2 = (not /d, h2, , . . hi. . . hpj) (/il not implemented) Er - (not h\, not A2,... not hi... nol hj*j) (no hypothesis implemented) Each situation (or image) Ek has an unknown probability n^ which we wouîd like to identify. Individual or conditional theoretical probabilities^, expressed as a function of FI^, can be associated with each isolated hypothesis hi. (1) Probability of h{

10. In somc cases, when it is préférable co speak of events rather than hypothèses, this number is rescricted to rive or six. 144

The SMIC meihod

The relation (1) expresses the fact that the probability of hypothesis i is the sum of the probabilities of the situations where h^ in fact occurs. (2) Probability of hj if h: is realized

where T(ijk) = 1 if /;;- and lu occur in E^ T(ijk) - 0 if h-t or h; do not occur in E^ In fact we hâve P* (i.j) - P* (Hj).P*(j) and the probability that h^ and h: occur simultaneously is equal to the sum of the probabilities of the situations where i or j both occur simultaneously. (3) Probabilities of i if not/

where Sftjk) = 1 when /(;- and not //.• are part of E^ S(ijk) = 0 when h^ and not /;.• are not part of E^. The necessarily valid conditions are:

Constraints (a), (b) and (c) are verified by tlie theoretical probabilities but not by the estimated probabilities: as a resuit the objective function that we propose to optimize consists of minimizing the différence beuveen tlie products P(i/j).P(j) which resuit from the estimâtes supplied by the experts and the theoreticai products P*(i!j).P*(j) which are expressed as a function of the 11^.

145

Reducing nncertainty: expert consensus ineihods

This is équivalent to seeking the probabilities (Yll} Yl2,.. • 11^ . . . n r ) of the 7' situadons possible, which minimize for example:

This is a standard quadratic minimization programming problem under linear constraints. Ai this stage it can be shown that there are multiple solutions for the n^ while the P* remain singular. A choice criterion is introduced: die solution taken to be optimal is diat which corresponds to the set of the n^ in such a way that the most likely scénario has the highest possible value, corresponding to 'reality', most of the experts bearing in mind that when they responded to the SMIC questionnaire this was a final image that they considered more likely dian any other. The solution corresponding to this criterion, Max (Max Iï^) 3 is obtained easiiy using the simplex aigorithm, since it is a linear function in II, to be optimîzed under linear constraints.

3 . 3 . R E 3 U LT S : M I E R A R C IIY O F S C E N A R I O S AND S E N S I T I V I T Y A N A LYS I S

For each expert the SMIC programme gives the probability séquence (II], n 2 , • • • FI,.) of the ?• scénarios which provide the highest value of the most probable scénario Max (Max 11^). We thus obtain a cardinal séquence of possible scénarios, which enables us to define the domain of the possible outcomes, retaining only those which hâve a non-zero probability.

Widiin die domain of the possible outcomes we can distinguish diose scénarios which are more probable than others, from which die refe146

The SMIC mclhod

rence situations and the contrasted situations can be chosen. From the probabilities of the scénarios vve can deduce the cohérent, simple and conditional, probabilities of the hypothèses, i,e. satisfying (a), (b) and (c). The rest of the method consists of sensitivity analysis to identify the driver or dominant variables and the dépendent variables. Sensilivily analysis This is a matter of estimating the variation APj of the probability P.- of the event j following a variation àP^ of the probabiîity P^ of the event i. The results take the form of an elasticity matrix: -

i

PG)

-- eij

The effects of an action on each hypothesis are measured by the coefficients of elasticity eij. We construct an elasticity matrix whose row totals give the sums of the relative induced in the probabilities of the other hypothèses for a relative probability variation of i (10 per cent for example). To some extent this sum translates the driving force of the hypothesis i on the others. Similarly the column totals allow an estimation of the degree of dependence of the hypothèses. Sensitivity analysis informs us as to which hypothèses should be enhanced or inhibited, thus orienu'ng the System in a desired direction. Thèse elasticides can be calcuiated by simulation (cycling the model of the relauonships several times through the probabilities). However, for technical and financiai reasons, it is established practice to measure the impact of one event on another by shifting the bar chart of the P(i), P(ijj) and P(ijnotj). Choice of the final images For each expert consulted, we now hâve a tist of the 2" images, ranged in decreasing order of probability. A list of some of thèse images (ten or so) can then be drawn up so that: • for each expert the sum of the probabilides of the images absent from the list is smatl; • for each image retained on the list there is at least one expert who assigns a large probability to it; 147

Redticing itnceriaiuty: expert consensus nmthods

we can then catculate the mean of the assigned probabilities for each image, obtaining a hierarchy of final images, and the corresponding scénarios. Base: Evoluîionary mechanisms of key variables Behaviour of main actors Key hypothèses for the future: SMIC inquiry

Set of most probable hypothèses

Scénarios: Pathways from the présent to final images FiG. 25. Domain of possible outeomes.

The référence scénario (a scénario often quoted and with a high mean probability) and contrasted scénarios are then chosen from among the available scénarios. The SMIC method can be summarized in essence as highlighting the most Iikely futures that will be dealt with by the scénario technique (see Figure 25), Once the final images hâve been worked out, the aim of the scénario method is to describe coherently the various pathways by which starting from the présent situation and inciuding evolutionary mechanisms and the behaviour of the actors analysed in the base - the future can be predicted.

3.4.

APPLICABÎLITY

AND

LIMITATIONS

Several criticîsms were levelled at the SMIC method, at Ieast as regards its first applications, notably: • A too limited mechanical application. • The objective function and the multiplicity of solutions. • The amalgamating of the responses of several experts. It is worth considering each of thèse points. 148

The SMIC mcihod

Limited and mechaniccd application

Every new instrument requires a learning period, during which time its use becomes an end in itself. Fortunately, for the SMIC méthode this was a short period and today it is only with great circumspection that it can be used within the framework of the scénarios method. The number of hypothèses or events that can be handled by tlie SMIC method is generally limited to six, this figure being chosen less for mathematical reasons than to limit the maximum number of questions that can reasonably be put to the experts. Objective funaion and the multiplicity of solutions

First, the objective function is in some ways arbitrary. Another function couid be chosen. Nevertheless, the one we hâve used has the advantage of conforming to the 'philosophy' of least squares. In effect, as in linear régression, we hâve a host of points (the unprocessed answers) and we seek the 'best fit' resuit that respects certain constraints (a straight line relation for the régression and the axioms of probability in our case). In any case, it is true that the optimum of the objective function is not unique and that there is an infinity of solutions for the probabilités of the k scénarios. In 1976 we suggested introducing a choice criterion and retainingas the optimal solution among the infinity of solutions - that which gave die highest value to the most likely scénario. The solution corresponding to this criterion, Max (Max lï^), can easily be obtained by using the simplex algorithm. This criterion has the advantage of partly removing the contradiction we hâve often noted among SMIC results^ which tends to assign a relativeiy low probability to the most likety scénario as well as to the initial viewpoint held by the expert being questioned, according to whom one or two scénarios were clearly more likely than others. Amalgamating responscs

Given computing costs, it is not possible to run as many passes of the SMIC method as there are experts polled; besides, there would be as many solutions as experts and arbitration problems would constantly arise. The solution we offer is to construct a typology of experts as a function of the proximity of their responses and to retain only those scénarios which seemed most probable to the majority of experts, grouped by type; i.e. as représentatives of a category. Given thèse limits, the usefulness of cross-impact methods such as SMIC remains considérable when it cornes to choosing the most likely 149

Rediicing imcertainiy; expert consensus melhods

scénarios that are reievant. They aiso help us to gain a better understanding of the stratégie behaviour of the actors driving the System by closely examining the image of the future that they generate. To better appreciate die technique's usefulness, we can now look at some examples.

4. Case-studies Around fifteen international surveys hâve been carried out using the SMIC method by mail and with very satisfactory response rates of the order of 25 to 30 per cent. The results hâve always provided an invaluable source of information for the relevant prospective studies. To iliustrate the SMIC method, we refer to two early applications to show the extent to which they hâve stood the test of time (which seems to us the least that can be expected of a prospective study). The first example concerns the probable trends in air transport in the Paris région to the year 1990 as they were envisaged in 1974 when a SEMA study was carried out for the Paris Airport Authority. Some areas of application o/SMIC over the pasi décade Long-haui passenger demartd to 1990-2000 Aircraft construction Air transport in the Paris région The French E3ost Office in 1990 World geopoliticai évolution France and world developments The conséquences of a îeft-wing victory in France in 1978 Firms in late 1978 The world petrochemical industry Offshore industries The European automobile industry The cosmetics industry Fairs and exhibitions in France The nuclear industry in the year 2000 Corporate activities and jobs to the year 2000

In this application, tliere was no inquiry involving outside experts, but a group opinion survey was undertaken by a working party which wished to test the cohérence of their views as they resulted from the exercise, together with certain assumptions about the 'most likely' scénarios. T h e issues in this prospective study were not trivial, since die 150

Case-studies

aim was to assess whether the u-affic outlook for the year 1990 required the construction of a third airport in the Paris région in addition to Roissy. In the cvent of a positive response it would hâve been necessary as of 1974 to acquire the land for this future development. Thinking back on events, it appears that the négative conclusion of the study was well founded. In any case the study clearly identified bottlenecks like passenger handling on the ground, average take-off delays, etc., the importance of which has since been confïrmed. However, the conclusions which turned out to be relevant relied partly on hypothèses that were not borne out by facts: for exampSe, the forecasts for économie growth were excessively optimistic (in 1974 the économie crisis was not generally considered likely to last). The second application relates to the global needs and prospects for nuclear energy to the year 2000, incorporating some results of the international survey published in Economia in 1977 and which, despite the passage of time, remain accurate today. By as eariy as 1977 it was clear that new forms of energy would not prove économie by the year 2000 compared with oil and nuclear power. It also seemed that the experts were divided about the development of nuclear power; some, particularly those outside France, maintained that nuclear power was only an intermediate step of minor importance, as an abundance of oil was expected: oil would be more abundant because of high priées (we were then on the eve of the second oil crisis)." Since then, history has shown that although the development of nuclear energy was necessary, it should hâve been carried out in a more flexible and less massive way because of the high degree of uncertainty about économie growth, the energy component of future growth, energy prices and conventional energy resources. In 1977, at any rate, very few experts believed in the development of breeder reactors. The décisions taken in France in 1984 with regard to that of Creys xVlalville confîrm the soundness of this opinion.

4.1.

AIR TRANSPORT SCÉNARIOS12

We examined those events which could affect traffic at Paris Airport for the period 1974-1990, an event being defined as the materialization of a trend to which a threshold is assigned, even if this trend is at présent only a potentiality. We identified six events which couid bc taken to characterize certain spécifie trends in air transport in the Paris région, \ 1. Thèse resuits ;irc akin to those quoted previousiy. 12. Study carried ont. by SEMA for the Paris Airport Authority; soc M. Godet (1975). 151

Reducing uncertainty: expert consensus tnelhods

The question asked was: Given that a System with six events can assume 26 = 64 states at the given time horizon, what states (or scénarios) are most likely, or most unlikeiy? The six events were: (e^J More than 50 million lourist-class passengers. This event characterizes an extrapolation of two trends: on the one hand, continued growth of passenger traffic and, on the other, a growing proportion of the tourist class in passenger traffic. (e^ An average ofover 150 passengers per aircraft movement. This event implies a certain load factor and aiso assumes increased and widespread use of wide-body aircraft. (e-^) An average take-off delay in excess o/20 minutes. This event includes a number of phenomena such as the saturation of airspace, which could be aggravated by the banning of night fiights. (eq) Air ticket priées decreasing by more than 3 per cent per annimi, in constant ternis. This means, for example, that if the price of an air ticket in 1974 is 100 francs, the same ticket would cost less than 63 francs in 1990, in 1974 francs. It could be said that if this particular event does not occur, air transport has little chance of becoming a mass transport option. (e^) French gross national product growing in volume at more than 4 per cent per annum. This event is favourable to air transport, to the extent that économie growth créâtes some business-class traffic and that a rise in real income increases the potential market for air traffic in gênerai. In the long term this event seems to be a limiting factor for the development of air traffic in the Paris région. Growth in traffic is often accompanied by urban development, notably around airport zones. This in turn threatens to aggravate environmental problems. (e^) Regulatory constraims involving a 20 per cent réduction in potential traffic movements. This could lead to limits being imposed on the total number of movements or even a ban on night movements.

[52

Case-sludies

Unprocessed data

Events up to 1990 are characterized as shown in Figures 26 and 27: (a) by the probability of occurrence of isolated events (Fig. 26); • ex More than 50 million tourist-class passengers in 1990

0.4

• e7 An average of more than 150 passengers per flight

0.7

•